AIFoundry.org

It is official folks! AI Plumbers Conference is coming to Minneapolis 11 Sep 2025 - 14 Sep 2025 as part of The Apache Software Foundation Community Over Code event. This is going to be our first ever event this side of the Atlantic! (or Pacific? I dunno... I get easily confused) https://lnkd.in/efxxBqnb

Now that #FOSDEM is disappearing from everybody's feeds until the next one, here is a longer read - behind the scenes of our #AIDevroom - a little bit of process, a little bit of doubts, a little bit of emotions (well... it was the first time for me being on the organizing side of FOSDEM, so a lot of emotions). If you are thinking about having you own devroom next year this read is for you. A bonus map of all the projects that participated at the end of the article. I'm sure one day the ecosystem of low-level OSS AI building blocks will grow and will be mapped by all the regular agencies. But this will remain the OG. Thank you, all the organizers who made it happen Roman Shaposhnik, Jarek Potiuk, JJ Asghar, Luca Corbucci, William Jones, Nina Davydova, Edith Puclla and everybody else who helped - Sergey Sergyenko and Steven Baker you were the best bouncers! Hope to have even better AI Devroom at FOSDEM'26!

#FOSDEM fatigue kicking in? Today in a more relaxed atmosphere we hang out with our fellow #AI Plumbers - https://lu.ma/fx9kupug ? Come join us to discuss the learnings from #FOSDEM, see some cool HW (thank you Tenstorrent and Felix LeClair) and do some unconference style group break outs!

Full house at our #AI #Devroom, no more seats and a line outside! I guess AI engineering is popular these days. Who knew? Matt Topol talk about Apache Arrow and storing tensors is right on spot, btw! Hurry up for next talks!

Right now in our #FOSDEM #AI #DevRoom https://lnkd.in/eDNr3iJ5 – Tanya Dadasheva & Iwan Kawrakow! You still have time to join us, please come!

#FOSDEM has already started! We are getting ready for our #AI #DevRoom - https://lnkd.in/eDNr3iJ5. Come and look for a guy in a green jacket – Roman Shaposhnik!

We got so many amazing proposals for the #AI #Devroom at #FOSDEM this year, that the selection committee decided to cut some of them shorter to 15-20 talks to fit more topics and speakers. That means that we won't have time for a regular Q&A, but instead we'll be collecting questions for speakers here https://lnkd.in/eAREwkRW and choosing 1 or 2 to be asked after the talk and we'll give the rest to the speakers for follow up in hallway discussions. You can also come to our fringe event on the 3rd of February https://lu.ma/fx9kupug and ask them there, a lot of speakers are attending. Feel free to start sending questions now and the form will be open during FOSDEM as well. Ask away!

Does scale of your AI matter? Scaling up always raises new challenges. E.g. if you run multiple nodes you have probably thought of a things like orchestration, monitoring and load balancing. We have a section about this in the AI Devroom at #FOSDEM. And since we really like #llamacpp universe, we couldn’t miss #Paddler. Come listen to Mateusz Charytoniuk to learn more!

Of course, talking about low-level AI engineering we couldn't forget about an important part of AI plumbing: HPC-like compute needs and libraries for AI inference. E.g. what format to use to store and move data in an efficient way in general is a good question, but for AI? Zero-reads, memory-mapped - where does it matter? Come to our #AI #Devroom at #FOSDEM and find out what the industry is standardizing on. Because let's be honest we got a few Apache Arrow related talk proposals, but no other alternatives. Thank you, Matt Topol for flying to Brussels from the States!

This is what we mean, when we say low-level #AI or AI plumbing! It still amazes me that things like this can be so open and don't have to be NVIDIA style black box. We have 5 HW related talks in our DevRoom 2:40-4:20, don't miss them!