Django in Space with Loes Crama

This episode of DjangoChat features Loes Crama.

Django in Space with Loes Crama
0:56:19
Published October 21, 2026
278 views

Loes Crama is a software engineer contributing to ESA's NewAthena mission at cosine measurement systems in Sassenheim, The Netherlands. We discuss her lifelong interest in space and her day-to-day work combining software engineering with data engineering and data science.

🔗 Links

📦 Projects

📚 Books

🎥 YouTube

Transcript

9,970 words · auto-generated · en Show

Automatically transcribed, so expect mistakes in names and technical terms.

0:00

Speaker 1: Welcome to another episode of Django Chat, podcast on the Django web framework. I'm Carlton Gibson. Joined as everybody Will Vincent. Hello, Will. Hey Carlton. Hello, Will. And today we've got a little clamor with us who built uses Django to build telescopes or something. Welcome aboard, Luce. How are you?

0:16

Speaker 2: Hi. Thank you. Yeah, I'm really good. Thanks.

0:19

Speaker 1: That's good. Go on. So we we've we weren't good I've s you know, I was talking about the telescopes, but you've got more to say. How how how How do you um let's google your backstory. Let's do the old the old Django chat

0:31

Speaker 2: backstory.

0:31

Speaker 1: Uh how do you find um yourself in the in the Django and Python world?

0:36

Speaker 2: So I was not familiar with Python, uh, and Django before I studied computer science. I had no b idea about both concepts basically. Um so when I did my computer science studies in university I got introduced to Python and uh I find it a a great language to work with initially just looking already at the syntax. was for me it was so easy for me to pick it up and it felt very natural for me to use in that sense. And um Uh Django was something that uh I got introduced to a little bit later uh actually in my career only at work Um so throughout university I did not do any Python projects there. There was just one course that was focused on uh

1:22

Speaker 2: on scripting. So we just wrote these short scripts basically uh to do some small calculations, for example, or read out the CSV file, stuff like this. Um but Django came a little bit later. Um

1:34

Speaker 1: So I guess I'd love to ask it as you know, so computer science, they often don't teach sort of what we think of as the practical end of programming, things like web development. Was there a web course even?

1:44

Speaker 2: So yes, there were uh a lot of them actually and we did a lot of practical uh projects uh a lot with actual companies as well. So we worked together with certain companies or or institutions where we would uh have a data set for example and build a web app around this. And this would mostly be done with web frameworks, uh things like Angular or something like this. And but the back end would usually be Something in Java like Spring Boot boot for example.

2:12

Speaker 1: Right. And then a JavaScript front

2:14

Speaker 2: end. Exactly, yeah, yeah. So not a lot of Python involved actually. I'm still surprised by this that it just hasn't picked up Uh really, that's great. For me the first one was uh Java and I feel like they they should have Done it a different way, first Python and then the rest, you know. So

2:38

Speaker 1: th that might be a um a time thing, right?

2:43

Speaker 2: Mm-hmm. Yeah, maybe and I feel like also Python is now more popular than like ten to twelve years ago. So maybe because of this There's also a little bit of a shift in universities as well where professors start to realize we should use Python instead to get people prepared for the workforce.

3:03

Speaker 1: Okay.

3:12

Speaker 3: They I've had this debate with professors and like, why don't you use Python? They're like, well, because I don't because I know Java and we and and then they say we want our students to know multiple languages. So You know, the younger professor can teach Python. Um and they certainly don't know JavaScript, most of them, right? Because if they're spending their career on algorithms like they don't So yeah, I remember here in the United States, MIT was one of the first universities to use Python for the intro class and was a really big deal five, ten years ago. I still think Carlton Java is you know, as popular if not more, still for the intro classes, largely because you get a senior professor who's just like, I just do it in Java and We'll get Python later.

3:57

Speaker 1: It's re it's real software engineering as well, right? Whereas

4:03

Speaker 2: the object-oriented part of course. Like it's a great introduction for object-oriented programming and how these concepts work, of course. So in that sense it's still a great language to start with, but nowadays Python is fully capable of also handling these types of concepts. So I think you can easily use Python as an introduction tool uh course as well. Did

4:25

Speaker 1: you you interested like you work you builds um telescope type, which we'll talk about in a minute, but did that Was there a connection there that drew you to computer science or

4:35

Speaker 2: um well I I was always uh when I was a kid I was already really into space and just anything related to space fascinated me. But I never realized it was actually like a viable career option. I just thought it's either an astronaut or nothing. And it was like, Well th the the chances of me becoming an astronaut It's very very small, so that's not going to be it. And um when I eventually had to choose uh and and pick a uh um um a direction in which I wanted to go for like university. I found out about computer science and I was like, Well, I've always really liked working with computers. and uh it was all also big interest of mine. So I figured I might as well just do this. And then later on I realized that, well, the space industry is actually an industry in which you can work

5:26

Speaker 2: not only up in space, but also there's a lot of people here that n that are there to build the actual satellites or uh use the use the data that comes back from a certain satellite or telescope. Like there's a lot of different jobs in this industry. And then I realized maybe there's also something for me in there. Um and it actually uh it actually opened my eyes a bit because I realized that there's quite a few software jobs in this industry as well. So that was perfect for me to kind of go into there through like a little backdoor or something. So it was not directly the initial um the initial plan to go into this field, but yeah, eventually I could still do it via

6:14

Speaker 2: software.

6:16

Speaker 1: Okay. And so you picked up Django at work, but then we know you from Django cons. Yes.

6:20

Speaker 2: Django Nukes. Yes, that's true. I um so I was introduced to Django I think a few years ago. I think it was like four or five years ago. Uh a colleague of mine had used it uh in the past and he pointed me towards it because I was looking for this new tool that uh could help us with this exact uh existing uh MySQL database that we had. And I was uh looking into how can we make this easier to use because it was a very old app and it hadn't been touched in a long time and I realized, well, if we still want to continue doing this and I'm now working here as a software engineer, I need to be able to maintain it and understand what is going on. And then I looked into a few tools. I was just uh doing this little research on what is out there, what can we use um as an interface to talk to this database.

7:11

Speaker 2: And then my colleague t told me about Django and um I Googled it. I followed the tutorial and within like one afternoon I had a running app, so for me that was perfect. I was just like, wow, it it works. So That's it. I I'm done.

7:27

Speaker 1: Right, there's the short for the F sites. With you know, an hour later I had it working and job done.

7:32

Speaker 3: Yeah, I I mean I think the key is you you already knew Python a bit. You had a formal education and you had a background in web frameworks, right? Because I think you know, as an educator, what I see is people have never used Python or they've never done a web framework. And so the polls tutorial I think is perfect if you have a little bit of experience as you did. But if you're a a total beginner, it's overwhelming. So maybe this is a Django on the med sprint. If I can come to Malta, Carlton, I want to finally push to get a you know, a second tutorial that's a little bit friendlier to beginners. Um

8:04

Speaker 1: tutorial for beginners. It's like when I was learning, I wanted to learn. I went um to I wanted to learn Python. I went to the Python tutorial. And the Python tutorial is really hard. It's for experienced programmers coming from another language. And I was like, no, this isn't for me. This isn't me. I found, you know, I found dive into Python stuff, which was much more my, you know Which are appropriate for.

8:25

Speaker 3: Wait, so um so the so the interface you built, were you then interacting through the admin? Did you build up like a website? Like what what was the next step uh on that project?

8:35

Speaker 2: When I just installed it, it was just the simple app that came out of the box, so just the admin and uh basically a direct connection to the database, no fancy APIs or anything in between Um and that just worked for us. And also because we don't have uh a really big user base, we decided to just stick with this and actually use the admin. This is still our primary interface, even after all those years. I think there's a lot of people who disagree with me and they say don't use the admin for your end users. I also highlighted this in my talk earlier in DjangoCon. And um there's just a lot of people who are against this whole idea of using the admin directly for the end users because you give them a lot of power.

9:22

Speaker 2: Um thankfully we only have about forty to fifty users because it mainly consists of the team that uh is using the app. So uh the people that use it actually work here. So if uh they have a question they can come up directly to me or other software engineers and also if If they uh do something that is not correct in the database, they they mess up some data, then it's just a quick conversation just saying, hey, you shouldn't do this in the future, and then we just move on. So Because we have direct contact with our end users, it's really easy to maintain all of it and also keep the data clean.

10:04

Speaker 1: Yeah. Th there's some line somewhere in the docs, I can't remember exactly what it says, but something like the admin is for trusted internal use. And it's like that kind of fits your your description.

10:14

Speaker 2: Mm-hmm. Exactly. Yeah. And I can imagine like if you have a Django app or you have external users that or maybe like forty thousand users that you don't know personally cannot contact directly, then of course you're not going to give them the Django admin interface because that's very scary and they can do anything at all times. So I can imagine for those use cases you would want to have some sort of interface in between. But for us it works really well.

10:41

Speaker 1: And do you use the the permissions to just sort of restrict what people can see to guide them into your own own yeah.

10:47

Speaker 2: Yeah, but the groups and the permissions, we use them a lot. Yeah.

10:51

Speaker 3: Have have you had to do custom things to speed up performance? Because the admin can be slow if you have a big database.

10:59

Speaker 2: Um, so the the database I would say it's not that big. Um I think we um We helped uh or we helped make it faster by using the uh the uh debug tool are the the tool that helped us give us more insight into which SQL Q queries are actually running uh in the back and there we had already done a lot of optimization to make sure that the actual admin pages load pretty quickly and uh this was already quite s sufficient for us

11:32

Speaker 3: Wow. And did you um are you using a theme or is it just the default theme?

11:37

Speaker 2: It's the default.

11:38

Speaker 3: Okay, because there's a lot of fun themes out there. I'm always

11:40

Speaker 2: Yeah, I know. I've been looking into it, but I'm scared that the that the users will be angry at me because they're like, No, we've done it for so many years like this and now it's suddenly pink or green, what's happening?

11:52

Speaker 1: They sound that they they make good uh veterans on the Django developer mailing list as well.

11:57

Speaker 3: Yeah. Yeah. Yeah. So so you mentioned in your your university studies you worked with some industries. How did you you're at a company Cosign now. Did you do a project with them as a student or how did you find your way uh to where you are now? would work.

12:12

Speaker 2: So well I always had well, the the idea of of doing something with the space industry in the back of my head and um At some point I um was following this minor for one semester, which was aerospace engineering and maintenance. And that's where I slowly got introduced in this whole world of aerospace and space as well. And during this semester we had some company visit uh to various uh space and aerospace companies, for example, uh Airbus. but also the European Space Agency. And yeah, this basically slowly introduced me into this world and uh I got to talk to a few people, I got to understand how software is also being used in these industries. And then I more and more started to realize that this was the way

12:58

Speaker 2: I want the direction that I wanted to go in. And um And eventually I needed to do a graduation internship and I heard from a friend of mine about cosine. Uh and it was also very uh close to my hometown and I figured well that's super easy, it's super convenient and it's exactly the type of work that I wanted to to do. So I just sent my C V asking if they had a uh internship for me and they said yes and then I did my internship here and after it I just basically stuck around and six years later I'm still here.

13:36

Speaker 3: So now you're running everything, right?

13:38

Speaker 2: Yes.

13:39

Speaker 1: Slowly work your way towards the corner

13:41

Speaker 3: office.

13:43

Speaker 2: Exactly.

13:44

Speaker 3: And is it is the main project this new Athena telescope? Is that the main project or are there other ones as well that you're working on there?

13:52

Speaker 2: So we have um two big teams here at Kosign. The one is uh remote sensing. And this is the business unit that focuses mostly on well remote sensing applications. So think of things like Earth observation where you send a satellite up to space and then you take pictures of the Earth, for example um to understand the climate or uh what is happening in in certain areas. And um the other team is Silicon Pore Optics. That is the team that I'm in. And uh This team is mostly focusing on um technologies that work with high energies, so in this case it's mostly X-ray energy. Uh and the biggest project project here indeed is New Athena

14:37

Speaker 2: And this is basically a contribution to the European Space Agency. And um we are going to make the the mirrors for this space telescope. Because it's an X-ray telescope, you cannot just create one mirror, but we need to make um hundred thousand individual mirror plates and we need to stack them in a certain way so that they're aligned correctly so that the uh x-rays can go through them and be reflected to the detector. So there's quite some some rocket science involved there. Um and uh the Django app is actually there to uh track all of these different components. So each of these individual mirror plates they ha all have this unique idea and we track them and we track their processes, we track the the quality, also the configuration, like how big is the plate, what type of material is it made of, these

15:30

Speaker 2: types of things. That's basically what we track in there.

15:35

Speaker 3: And then is that and that's going live in 2028? Is that the plan? Something like that?

15:40

Speaker 2: Uh 38, I believe.

15:42

Speaker 3: Oh, 38. Oh wow.

15:44

Speaker 2: Yeah. So we um We go into production in like a year or so, then we actually have to deliver the mirrors to Issa. Now we're still in this sort of like ramp up to uh production. So we still have some time to do some R and D and test a few things and experiments. Um this is done d on a daily basis in the in the labs that the uh the scientists and engineers work in in the clean rooms and they basically yeah try to to find the best configurations for all these mirrors and once we actually go into production then we deliver them to Issa.

16:22

Speaker 1: You you talked about this in um uh where was it, Athens? You talked about this in Athens, um in your DjangoCon talk. Um you make It was this wonderful um story of like how you've got like this ten year timeline, but also it's quite tight. It sounds like 'cause it sounds like the the programmer's worst nightmare is that you've got far too long before you have to deliver, so you won't do anything for nine and a half years and then really rush. But There's none of that, right?

16:46

Speaker 2: No, no. And thankfully there's some sort of like um milestones that you work towards. So we have contracts with Isa don't that don't span like twenty years but only like a year or maybe three years. So there's still some reviews in between and some things that we discussed that are deliverables in the meantime. So yeah, we actually have some things to work on. It's not that we can just sit back for nineteen years and then in one year we just Work really hard and hope for the best.

17:15

Speaker 1: Yeah, so so any take home man um project management tips for the rest of us who don't have

17:24

Speaker 3: Yeah. Uh I'm curious what um what does a day look like, right? Is is there a typical day? Is every day different, right? That's always you know, I know people who are outside of the industry are always They don't really know what a programmer does. And partly it it varies so often. So if there was a typical day, what would that sort of look like for you in terms of being in the office and and what you're doing?

17:46

Speaker 2: Um well I wouldn't even know either because every day is very different for me. Um so because my my job is not only software engineering but I do some data engineering here and there, some data science if I'm interested in in it, if I feel like it. Um thankfully I get a lot of freedom to work on the things I that I actually want. So this is really nice. Um so it kind of also depends on um we look at basically the the different open issues that are there in in our internal GitLab instance. Um we have a look there, we see if Are there any issues that have a really high priority? Are these things urgent? Should they be done immediately? Then of course uh they will go first. But if that's not the case, then it usually depends on whatever's there

18:32

Speaker 2: still left to do on on my to-do list and it could range from setting up uh some some data pipeline, for example, um with some data that we collected from the clean room that needs to be processed and and plotted and visualized in some way. But it could also indeed be something like uh standard Django maintenance uh on some form on or maybe on the admin side, maybe a new feature that the user wants to have implemented Or it's something that is more about uh things like data engineering, like how um How are the data pipelines actually running? Is everything running correctly? And uh what are the logs saying?

19:18

Speaker 2: Stuff like this, more in terms of observability. So it it ranges very much. Um but I really like this. It's sometimes a little bit of a challenge, but it's really fun. Yeah.

19:27

Speaker 3: I mean that sounds like a lot of a yeah, a lot of autonomy, a lot of areas you get to to work in. I mean I think in um DjangoCon Dublin, you came up to me, I think uh I gave a talk on Django for AI and one of my questions was how many people are doing web uh work versus data science work and albo almost nobody's doing both professionally But you are, actually. Right? So you you're which is um so you should give an updated version of that talk, right? But no, that sounds w I mean that sounds like a perfect setup for an engineer to have

20:01

Speaker 2: Mm-hmm.

20:01

Speaker 3: New things to learn and not just the same old stuff. That sounds like

20:06

Speaker 2: it's nice that you can uh learn a few new things and not just doing all the same things day by day but you get to work with newer technologies and and try them out first. See if they actually work and if they don't work then they don't stick around and that's also fine because we do want to have our tools very stable, very reliable because it's such a long-term project. We really want to make sure that this would still work in ten years. And um And because of this, we do really test our tools first before we actually decide to go on with them. But I think Django is a perfect example uh of one of these tools because it is it is so stable and even over the different major releases it it still proved to be uh so well done and

20:52

Speaker 2: it still works the way we want to. And uh I remember when I was just getting into Django was I was reading the book Two Scoops of Django and we had a really old edition here um in the library at work And I think it was one of um Django one point something. And at that time we were already in Django four point something. And I was reading it, but still everything made s sense because the core of Django did not change, basically. Uh I could still use the book for a lot of things and of course there were some minor things that may have changed or some name changes maybe here and there and I had to look it up in the documentation, but the core of Django was still the same. And that's exactly what we're looking for in terms of software tools for these types of projects.

21:38

Speaker 1: Yeah. I think that the people using Django in the real world, that's a story that we hear again and again and again. It's that it's s it's it's s um stability, it's reliability, it's the fact that it doesn't change that actually why companies choose it and what you know. as you say, a tenure, ten-year, twenty-year project, you know, the framework that's going to be with you from start to finish is of value, right?

22:00

Speaker 2: Exactly. Yeah.

22:02

Speaker 3: I think that was um I had all the versions of two scoops. I think it was 111, I think was the last one, and and then there was some stuff around the URLs changed in 2. 0 and but it was Um, Encyclopedic, the book, really you could dive into sections, so it it worked really well to have great advice on it.

22:21

Speaker 2: Exactly. Um I

22:22

Speaker 3: I did want to ask um so databases. So you mentioned there was a MySQL database. Do you is it all MySQL? Do you sometimes use Postgres? Do you have needs for non-relational databases? Like because the data is varies, I imagine, right?

22:38

Speaker 2: Yes. Um so for the the components themselves that actually go into the telescope, we can just use MySQL for this. uh it's just because there's also di uh various relationships between these components like also physic physically it makes sense to have a relational um database as well Uh and that works really well for us. But there's indeed a lot of measurement data and actual experiments that are done uh where data sets roll out and we store them in various different ways. So the the biggest um other data source that we have is MongoDB and this is basically well non uh SQL uh data of course and it's very flexible for us

23:23

Speaker 2: because If a experiment changes over time, then that's okay. Like we can change the the key value pairs and nothing will be affected in terms of previous um measurements or records. So that's really great for us. And we store all kinds of data sets there, uh short-lived experiments also in case we have something that we just want to test out for let's say like two weeks and then we decide not to continue with this method. Then it's okay, we still have a record of the data at least, but yeah, we can store it and we don't have to think about any structures or or tables or anything like this. So the uh the MongoDB database is is used a lot uh in this case. And then we also have uh InfluxDB.

24:09

Speaker 2: which is a um a time series database which we mostly use for environmental data. So we monitor, for example, the the the air quality, the temperature, humidity of the clean rooms, uh this type of data is all stored in Influx T B So yeah, these are basically the two other fake data sources that we use here.

24:30

Speaker 3: Yeah. Well that that's great to hear because MongoDB had added official support for Django, I think two years ago. Um Yeah, because they saw in the survey results a lot of people are using it, and so uh it makes sense that another example of where it's being used.

24:49

Speaker 2: Yeah, no, I've yet to try out the the the connection between Mongo and Django. I've not tried that yet, um because most of the data that we have in Mongo we read out through uh Jupiter notebooks and some custom um uh aggregators that we've written in in Python and then use that in uh in Jupiter. But yeah, I think there there could be uh potential to uh To connect to the job. Another Django admin there. Django. Exactly. Just the admin works fine.

25:24

Speaker 1: Just the inspect the eBay. Um exact so you work in this very conservative um environment, but what about the new the new totless, the new the new tools, the the AI agent and assistance and and things like that. Are you are you giving those a go? Are they do they find a space?

25:42

Speaker 2: Uh yes actually. Um we see that everyone in the company is interested in a AI is just because people are curious about new technologies in general. But we also realize that we need to protect like in intellectual property and we uh focus heavily on privacy and we try to limit the tools that we have uh uh in the cloud basically so we make sure to host uh as many software tools local as possible.

26:11

Speaker 1: So we have to self-host GitLab.

26:13

Speaker 2: Exactly, yes, yes. And we don't use tools like Teams or Slack or something like this. We have uh

26:20

Speaker 1: elements.

26:24

Speaker 2: Yeah, so everything is very local and it's the same for AI because at some point people started to use ChatGPT, Claude, everything looked super interesting, of course, and it was like the shiny new tool that everyone wanted to use So um yeah, within software we realized that there was still a need to use these kinds of tools. But at the same time we didn't want to have people upload whole design documents of of telescopes to some random server in in America. So this is uh where we uh looked into a local AI stack and um we basically set up this local uh AI tool. that consists of OLAMA. Um and we use the Lama models in here as well. And we have open web UI where we have this uh chat interface so it feels very similar, very natural for the end user, similar to for example in Chat GPT where you can just type in

27:17

Speaker 2: your question. And we have also some custom scripts where we upload the different documents to an internal document database. So uh in this in this way we still have our uh we still have the context that is required uh to understand certain problems or understand certain projects Um and the local AI is actually really, really good at understanding these di different things. So we have like technical notes or maybe internal minutes from meetings and with the help of these documents we Yeah, we get quite far with this local tool.

28:00

Speaker 1: Okay, that's super. Can I are you starting to technical question about the size of the Sorry Uh size of how how how big a machine do you need to run, you know?

28:09

Speaker 2: So we actually recently had some hardware uh upgrades for this because yeah, it was indeed required because otherwise we couldn't upload as many doc documents and we couldn't run heavier models. Uh I'm not sure about the exact specs, but um yeah the upgrade has helped significantly.

28:30

Speaker 3: 'Cause one one of the conversations at DjangoCon US was around so some people have the new MacBooks, the M5s, where they can run a lot. But then a number of the consultancies are buying, you know, foundation or NVIDIA chips like literally in a closet for their team. Um so I guess that was my question. If you know, is it like In a closet somewhere you have some fancy laptops or do you have your own private cloud or do you know how the that's being set up?

28:57

Speaker 2: So uh yeah, we actually have a lot of servers just here. Uh just physically here, so it's it's all running

29:05

Speaker 3: crazy. It's so like you're you're too young, but for Carlton and me, right? Like I was around before the cloud was a thing. And it was it was literally in a closet. And there was like one guy who was grumpy. Carlton, you would have been good at this role, right? It was this like this grumpy guy who like nobody wanted like was the only one who knew how to do anything. And you And that was how it all worked. And then everything went to the cloud and now, yeah, everything's coming back. And it's right, c it's interesting to see To see that, Carlton. Um

29:35

Speaker 1: well I I used to used telling that story when I was doing my PhD, I used to sit in the this downstairs basement room in the philosophy department and read because there was a server cupboard in there and it was really warm. Who is like the warmest room in the debunk?

29:48

Speaker 3: In damp Britain, yeah.

29:50

Speaker 1: Yeah, yeah. No, it was brilliant I'm sure there was a question for Lucy there.

29:55

Speaker 3: No, well I think um

29:57

Speaker 2: No, it's a great story.

29:58

Speaker 3: I mean Carl you you said A AI first, Carlton I I held off, so now I can say pie charm. So We are adding JetBrains is adding a bunch of functionality for local, you know, in addition to cloud support for frontier models, local things that you can run uh on your device and I think it's yeah, it's already pretty close. And the the divide that I see actually was people who have access to frontier models. They talk about how the the tip of the s the the you know fable and the newest models think too much. So you actually, you know, you give them the planning, but the implementation, you use a in quotes dumber model that won't, you know, lie to you and stuff. So I I think Yeah, local all the way, you know, for cost reasons and security reasons.

30:45

Speaker 3: Um I'm surprised more people aren't aware how bad the security is. Um I mean, even in the United States, there's court cases. I have friends who are lawyers who are now using AI and they don't think about the fact Once it's in ChatGPT or Claude, like it's there. Um so I suspect they're just a couple years behind everyone else in realizing like it's the same as typing something into Google. Like if you type something into these services. Not to mention they'll steal it and then, you know, then t win a fields medal or whatever, a Nobel Prize and and cut you out, um, which they also like they also seem to like to do. Um do you uh uh well one actual question. Do you have a f uh a favorite model for for coding that you've seen? I mean I think in the notes you mentioned Quin3

31:30

Speaker 3: or Gemma. Is there one that you've

31:39

Speaker 2: Especially for Django, I would say. And uh some uh yeah just generic setup uh things like configuration files, think of like a GitLab YAML or stuff like this. I've been using it in in in that area lately uh a bit. But um yeah. I'm

31:58

Speaker 3: But that's what I hear too. Like I ran a open space at DjangoCon basically like let's let's compare models and you know, people complain but a lot of people s liked Quinn III in in particular. I think in it would just do it. Again, these people would talk about in a anthropic context, they would say fable for reasoning, but opus for for doing and um sorry sorry one more question on this um do you are is your team using custom harnesses at all or are you just interacting directly with the the models

32:32

Speaker 2: Uh just interacting directly, mostly. Yeah. And uh most of the users just make use of Open Web UI directly and then just asking a question or two, um nothing too fancy or anything.

32:47

Speaker 1: And how do you think it changes the role? Like, you know, does it it because do you think it puts us out of work, you know?

32:56

Speaker 2: Um no, definitely not in this field. Because I've tried it with data analysis and it's it's horrible. It just doesn't work. It's it's not able to comprehend what we're doing. We're we're talking like things like astrophysics or Mathematics it's just not able to to pick up these things and it will spit out something that's just incorrect. So in this in this field I think we're fine. Um, but I have to say with Django it is really good, it's really accurate. Um, also probably because there's just so much out there on the internet, so it's been trains uh on all these different uh sources. But I'm not worried uh that it will like do the full on takeover or anything.

33:41

Speaker 2: There's there's still so much uh that humans do better than machines and Uh, I think humans are way more critical of the work they deliver and an AI doesn't do that. It's uh it just spits something out and it it doesn't review it and have like a double check and see if it's actually the good way to do something, a certain task, for example. And humans actually are very critical, especially on themselves or on other developers, so yeah.

34:15

Speaker 3: I don't wanna B2 on the AI side, but I think harnesses, if you look into using custom harnesses with uh AI models, they do do a lot of the review and and checking and so You know, for your team, I would strongly suggest looking into some harnesses. Um it makes a big difference, actually. I I I still agree. There's still something humans can do better and you know, judgment and stuff, but Um if you're just d if you're not using a harness, I think you're really missing out, especially with Django on what what they can do. And there are open source local harnesses you can use.

34:52

Speaker 2: Yeah, no, it's definitely worth checking out indeed. Yeah.

34:56

Speaker 1: So I want to come back because you said you built your whole wrap around the admin and then you gave it to 40 users and they build it and you're five years in. You must have done some pretty gnarly customizations.

35:07

Speaker 2: Yes. There's some things that slipped in between the admin classes sometimes, but It's it's still it's not hacky or anything. Because it is so

35:20

Speaker 1: you write it. Exactly.

35:22

Speaker 2: I understand it. It's No, I think uh it it doesn't it doesn't feel hacky at all. Um and I think the the admin actually allows for a lot of customization and because of this it it feels natural to add things to it and and Uh there 's a lot of extra packages that also make the admin easier to use. Um there's for example one package that is called Django QL and you can do SQL like queries on the data um in the in the search bar, in the top search bar in the admin. And it w works really well for some of the u users who want to uh filter on results that are a little bit more complicated than just the generic lookup something and name like uh a part of a name or a description.

36:10

Speaker 2: Um they can look on multiple conditions, for example, and um So these things are are really nice additions to the uh the default Django uh admin. And I it it doesn't feel like it it shouldn't be there or anything.

36:26

Speaker 3: So I uh I l I love this about Django. Like so Django QL, I I hadn't heard of it. I go look it up as as you're speaking, and you know, there's over a thousand stars. And so You know, I just love that I I every day I learn something new about Django, right? It's big enough that there's and clearly it's, you know, nine years old and you know, like or yeah, nine years ago it started, so there's just such a depth to

36:51

Speaker 2: Yeah.

36:52

Speaker 3: Uh to things. But I um I know you there's a number of other tools that you use. I think uh Django Simple History is that one as well? Is are there some are there some third party packages that you really can't live without with your work?

37:04

Speaker 2: Well indeed the the the Django simple history that's indeed the one that we use quite a lot because we just really want to monitor and and track all of the history. to all of the different components uh over time. We want to see what has happened to all these different things to uh ensure that all the procedures of course went right and then we can actually uh validate but also show isa for example like see this is how we made the mirror and exact it's exactly step by step how the procedure says we should do it. Um and also we make use of Django Leaflet. And this was kind of something that I was just personally interested in. Um Django Leaflet is basically uh a package that helps you uh show different maps on uh

37:49

Speaker 2: in the admin and um make use of OpenStreetMaps, for example. And we also track the the geolocations of each individual component because we sometimes ship the the plates to another facility where we do certain tests. So it goes to Berlin, for example, or it goes to Barcelona. And it is just really interesting for us to see some kind of like the the history of where all these different plates, uh these mirror plates went to and um we even mapped it out in such detail that we have this geospatial map of all the different clean rooms and we track for each different plate in which clean room they are currently at

38:35

Speaker 2: and where they're going. So together with the simple history we can look up the Django uh locations, uh the historical locations And then we can exactly see where a a mirror plate has gone over time. And this is super interesting. And it also helps for clean room operators and coordinators to see Are we actually doing things right and is everyone tracking everything according to the procedures and how it's actually set in the documents?

39:10

Speaker 3: And that's all Django Rest framework uh for the APIs.

39:16

Speaker 2: Sorry, I

39:17

Speaker 3: Is it is it Django Rest framework? Is that what you're using for the APIs?

39:21

Speaker 2: Yes. Um so we basically use this uh this Django app as like a main yeah central point of truth where we uh look up everything from the current state of of what is happening right now in the the clean room to well, the history from maybe two years ago or even further back. So because of this it's also important that we can retrieve this information through indeed an API, um and not only just through an admin interface. So uh with Django Res framework we've uh we've made this small API that can just retrieve all of this information and we then use this API again in our some of our Python tools To fetch the data and to see, for example, the latest status of

40:10

Speaker 2: some plates uh to understand what's happening to them or get a filtered list of certain components, stuff like this.

40:21

Speaker 1: You talked about um Django QL and being able to type um slightly more complex uh SQL expressions into the search to kind of filter down. It sounds as if your data model would be one of those ones where access to raw SQL would be quite important because complex queries and It's not always so easy to construct the right filter of values combinations from the ORM. How do you find the the the the expressiveness of the ORM? Is it sufficient for m for a lot or Do you need to drop down to raw SQL a lot?

40:56

Speaker 2: No, I think it's actually sufficient for us. And I think uh that's also thanks to Django QL. Uh a lot of users are now happy with that package. So there's not a lot more extra to add. Uh the current setup w that we have is sufficient for for me as a developer, but also for the end users who just want to look something up with some extra filtering or sorting options.

41:19

Speaker 1: Okay.

41:22

Speaker 3: Um before we get to projects and books, I know there's some listeners who are interested in the space industry. Like what sort of advice do you have for someone who maybe knows their way around Python and Django? But thinks like what do they need to do or or think about to to also work in a space industry?

41:38

Speaker 2: Well, I think um You shouldn't be intimidated by the industry because I was at first definitely intimidated by it because I thought, Oh, I need like a physics degree or I I need to have a master's or a PhD in astrophysics and I need to be able to name every individual black hole from from the top of my head or something. But that's not the case. Uh space is a very, very uh big industry and there's a lot of different roles and um because of this it's also very accessible for software engineers. Uh I think a lot of the people that work in his industry they come from this sort of like intrinsic passion and uh their their own obsession with it

42:24

Speaker 2: basically. So every time you talk to any anyone who's in the space industry, they have some personal relationship with space, basically. who maybe has done a full PhD on on one specific uh black hole and is super obsessed with it while the other person just is really fascinated by space and watches space documentaries or or even things like Star Trek every night. It's it doesn't have to be like super complex or anything. As long as you're interested in the field. I think you can get quite far and if if you're curious then that can already help you uh get really far here.

43:07

Speaker 3: I I think I'll have to point my my six-year-old son in this direction because on the walk to school this morning he was asking me about black holes. Like he's upset. He's like, what's inside a black hole? And then he said, Well what happened before the Big Bang? And you know, so maybe maybe he's a philosophy person too, Carlton, right? He's he's fascinated and he's like, What do you mean you don't know? You're an adult, you have everything. Or you know you know everything. But yeah, that that's great advice. Not the first. Not the first. Um Carlton, did you have any add additional questions before we

43:44

Speaker 1: No, not real. I mean, you know, we could I could talk forever about you know the the the the work and the I I'm I 'm not sure.

43:54

Speaker 3: So you've been to three, I think, right? So Vigo , Dublin, and then Spoke in Athens. Um I I mean I met you in Dublin. You're You're not shy about going up to people and talking to them. Is that natural or did you have to cultivate that, right? 'Cause sometimes like for me, my first DjangoCon, I was a little more reserved and it was really only Carlton at my talk saying something and then talking after that sort of opens it up. But like, you know, you just came up to me after my talk, which I was like, oh that's great. So so I guess the question is like, did do are you have you always been that way or is that something you built up over time? Because if you're an outsider, you can maybe feel like, oh, it's you don't know anyone at the first event.

44:36

Speaker 2: Well that's true because indeed when I first went to Vigo I didn't know anyone. So it was just me on my own

44:42

Speaker 3: and yeah.

44:42

Speaker 2: I didn't know what to expect like from Django Con in general, but I think the Django community is also very welcoming And it's one of the few communities where you instantly feel comfortable and things things are not weird in in the sense that if you go up to a stranger and you ask them a question like, hey, what do you do? or how do you implement Django in your in your day-to-day life People are not uh not shy to give you an answer and they don't judge you for not understanding uh certain parts. If there's, for example, a package that you've never used before, then they're like, oh well, we should have a look and and read into it 'cause it's a really nice one or uh people are able to explain complex uh topics. So I feel like it's partly the

45:28

Speaker 2: the community that uh just makes it really easy to yeah, talk to people and just go up to them. And also I feel like if you go to a conference, I'm I'm just a curious person. I just want to know what are you doing? What what what keeps you busy? I I wanna know. I'm just very nosy.

45:50

Speaker 1: I remember you from Vigo from the sprints when you came up and you're like, You you do Django Filter and I'm like, Yes and you're like, Can we do this to say something or something? I was like, yes we can.

46:00

Speaker 3: Well I I think I mean the older I get, I think that's Well, the the two attributes I would recommend to anyone is obviously kindness is number one, but number two is cre is curiosity. Because it just takes you on so many interesting places and things and I think when I was younger I sometimes felt like you see people who are maybe quicker in math or, you know, faster in something, but if they don't have Curiosity, like that's what you need for the long term. Um, and so along with kindness, those two really win out over time. Um Yeah, so exactly. Like I'm I'm the same way. I'm always like, well, I don't know, how does that work? Like tell me more, you know?

46:41

Speaker 2: Yeah.

46:42

Speaker 3: Like um

46:43

Speaker 2: Yeah, exactly. It's just a matter of wanting to know and wanting to learn, I guess.

46:48

Speaker 3: Yeah. Yeah. Um projects. So I'm going to switch. So sometimes we recommend projects. You've mentioned a couple. I think there was one we have in the notes that you want a non-Django one you wanted to call out. Is that right? Marimo?

47:01

Speaker 2: It was actually one that I was introduced to uh a Django Con in Dublin, I believe, two years ago. Um it's called Marimo and it's kind of similar to Jupyter notebooks, but it's more dynamic. So you can also write your own notebooks and play around with different data sets, but it has a little bit more uh dynamic uh elements to it. So for example already in the UI there's some some sliders and some filters for example and you have immediate response in in your plots and I think it just looks very neat and very clean. Um, we do use Jupiter at work, but I sometimes still play around with Marimo just because I'm curious to see how I can visualize the same data in

47:46

Speaker 2: this framework instead.

47:48

Speaker 3: Yeah.

47:48

Speaker 1: So it's a bit of a

47:49

Speaker 3: it it has a lot of momentum. Yeah, more modern. Yeah. PyCharm's added support, so I've been I've been pulled into discussions around it and it looks very, very cool.

47:58

Speaker 2: Yeah.

47:59

Speaker 3: And is gr is po very popular.

48:01

Speaker 2: Yeah, and it's and it's getting more support with other uh packages as well in more of the the scientific Python community as well. So that's really good to see as well that it's it's not just this hype project that dies out after a few months, but it's actually it's there to stay. Maybe.

48:22

Speaker 3: Carlton, do you want to go next? Do you have

48:24

Speaker 1: Yeah I had a project which just um is Django S E S, which is um a Django email backend for Amazon's simple email server. So I I use that and I've always just used the um the API for it and whatnot. But this package looks like it's got it's a nice email backend Um I've always used the S SMTP to actually send it the messages. But um this uses the API directly. And then what it's also got though is the simple notification service. webhook integration so you can get the the notifications for whether your emails are delivered or where they bounce and that bit of it you've always had to wire up yourself. So um I s this came across my radar. I'm like, oh, I might be able to delete some code by adopting this. So I'm currently digging into that. Um

49:06

Speaker 3: Well, we we didn't um talk about this in advance, but my recommendation is actually also on this. It's uh Django any mail, and I have a story around it. So partly Django 6. 1 updated the mailers in the settings and um I've been busy doing that with my uh new version of Django for beginners which will be out by the time this podcast comes out. Um but Django any mail is a tool I like to use and I use I used to use Sendgrid for my personal projects. And I was updating the 6. 1 and I saw, oh, there's no support for so Django and email is my package. And then s my story is Sengrid uh didn't have support didn't allow the Django Animal volunteers to do the testing that they wanted to do.

49:52

Speaker 3: And so for a year and a half this was sort of stuck. Um and then uh I and some other people kicked up a little bit of a sp hissy fit online about this and we're able to connect with the right person. So I'm happy to say that as of uh yesterday There's now support again, and so Django NEML has official support for SendGrid as well as many other providers, uh including Mailtrap, who I actually ended up switching to because they sponsor Uh Django and email. So a long-winded way of saying like these these tools are important and useful and it's beneficial to companies to both respond to them and also to support them. Um, but I'm happy that had a nice conclusion. Um so thank thank you to the people at Sungrid who who ended up uh fixing that eventually.

50:44

Speaker 3: Yeah, yeah, I know. I mean I I said something and I complained to Jeff Triplett and then Frank Frank Wiles um said something and then someone else said something. So it certainly wasn't just me, but And then um oh and um oh uh yeah someone online um who is it at uh at Wagtail um anyways got connected to the right person at SunGrid. Um so yes, it all works out. Um, books, the fun part. Uh we do book recommendations. Luce, do you have a book that you wanted to recommend?

51:15

Speaker 2: Uh yes, there's one that I recently read, uh and it's called Weapons of Math Destruction by Kathy O'Neill. I think that the title in itself is already genius. And it's basically how some algorithms that are used in big data can actually cause more inequality and eventually even threaten democracy um it's it can have quite a big negative impact on on certain groups in a population. And there's some really nice uh

52:10

Speaker 3: I l I love that one. I think it's a it's a few years old. It was before AI or LLMs became big. Um when I think the author was working inside I forget if it was Twitter or one you know, one of these companies and seeing firsthand how it perpetuates stereotypes and all again, all these things that are a little more obvious today, but were less so back then.

52:32

Speaker 2: Especially reading it in this day of age, y you see that there's some things that she had predicted and you're like, wait, these things actually came through and it's kind of scary but also really intriguing to see. Yeah, it's a really good book.

52:46

Speaker 3: Carlton, another philosophy book?

52:49

Speaker 1: No, I'm still reading Concept of Mind by Gilbert Ryle that I was reading last time when we had uh John on. Um philosophy books they go slowly, but it has made me pick up a bit of Aristotle, I've picked up a bit of Kant, I've got picked up a bit of Nelson Goodman, so I'm I'm just having massive fun. Um when I finished it I'll have a fresh book recommendation. But in the meantime, concept of mine, Gilbert Bo Gilbert Rutt Classic of twentieth century British analytic philosophy.

53:12

Speaker 3: Yeah. It was new to me when you mentioned it before, so I looked it up and um

53:17

Speaker 1: Yeah, it's I mean, you know, it's it it's it's to my taste. Uh

53:21

Speaker 3: I think it's above my head, but yeah, but it's inter the the idea is interesting. Um I'm gonna read this a little more pedestrian. So I've read this, but um this is called Slouching Towards Utopia So I I like economics and history. This is a long book, but it's basically the history of ca uh economics in the 20th century in the United States and Western Europe. And Seems like it might be dry, but i it I think it's really interesting and it puts all of the historical events in an economic context context. Because of course, money kinda drives a lot of things. Um, you know, like when did the United States get into World War Two? It wasn't just to help out, it was also because the you know US banks had lots of loans. Um but he talks quite a bit about inequality and all these K-shaped recovery things

54:08

Speaker 3: and um so I think a little bit like um Capital the Piketty book, but a little more readable, I would say. Um so anyways, I if I was gonna recommend a single economics book, I would say that. And then there's a webcomics book, economics, that Carlton actually has has looked at Um, I'll put that in the show notes. Those are my two favorite economic history books for sure. Um What else? I've all we've done is ask you questions. Is there any topics we didn't ask you about or that you wanted to mention while you're while you have a microphone?

54:42

Speaker 2: Um no, I think um I think you asked me most of what I wanted to to say today, so Yeah.

54:52

Speaker 3: Okay. And then y so you're you're online, you is it on Medium where you're still doing your writing, or what's the best way for someone to learn more about what you're doing?

55:01

Speaker 2: I think LinkedIn, that's where I'm most active. Yeah. I I do still have a medium account, but I don't post as often. I I feel like I should again, but LinkedIn would be the the main channel.

55:12

Speaker 1: It's the internal guilt of the blogger.

55:15

Speaker 3: Yes. And your talk, I think the videos are are not up yet as when I checked, but as soon as they are we'll we'll add them. Yeah. So Yeah, I was I was sad I couldn't see it in person. But I know Carlton saw it and I I read many reviews of it.

55:41

Speaker 1: volunteer effort doesn't always stretch to speedy turnarounds, so that's fine. When they come.

55:47

Speaker 3: Um, well thank you so much for taking the time. Um And thanks for, you know, being part of the Django community both, right? At conferences, giving talks and now you know, now you're one of the people when you go to conferences, people come up to you and say, Oh, I saw you gave a talk or I I heard you on the podcast. So now you can be in the the other side of that uh conversation.

56:08

Speaker 2: Yeah, well thanks for having me today. I really enjoyed it.

56:11

Speaker 1: It was really fun. So joining us.

56:14

Speaker 3: So Django Chat DjangoChat dot com and we'll see everyone next time. Bye bye.

Presenters

Podcast

DjangoChat

Note: We understand that names change, people change, and bodies change. We respect each individual's journey and privacy. If you have any concerns about a video or need us to remove content, please don't hesitate to contact us. We will handle your request with care and promptly address any issues.

More videos by Loes Crama

More videos from DjangoChat