Francisco Javier Arceo
banner
franciscojarceo.bsky.social
Francisco Javier Arceo
@franciscojarceo.bsky.social
180 followers 100 following 27 posts
Open source AI @ Red Hat | Maintainer of Feast (the open source feature store) | Ex Affirm, Fast, Goldman Sachs | Columbia alum | My silly opinions
Posts Media Videos Starter Packs
Reposted by Francisco Javier Arceo
🚀 Exciting news from the Kubeflow community!

Welcome Francisco Javier Arceo & Julius von Kohout to the Kubeflow Steering Committee! 🎉 Huge thanks to Mathew Wicks, Josh Bottum, & James Wu for their leadership & dedication. More information: groups.google.com/g/kubeflow-d...
#Kubeflow #OpenSource #AI
Welcome to Our New Kubeflow Steering Committee Members
groups.google.com
I’ve been elected to the @kubefloworg.bsky.social Steering Committee! 🎉

Kubeflow is the open-source AI platform that makes AI/ML simple, portable, and scalable that was originally introduced by @googlesearch.bsky.social.

If you're building in AI, let’s connect! 💪
Would love folks to engage with the @kubefloworg.bsky.social community.

We’ve been building AI/ML infra for a long time and welcome new folks!
Reposted by Francisco Javier Arceo
The reason I think it's important to spread the message that using modern AI tools is difficult is that I don't want people being embarrassed to ask questions about how to use them - or quietly assuming they "aren't smart enough" and giving up
General consensus in the replies and quotes of this seems to be that the entire concept of "AI skills" is a joke - how hard is typing text into a chatbot, really?

I will continue to argue that it's genuinely difficult, and that the challenge in using these tools is widely underestimated
No, they're not caught between anything. Stop it. The fact that AI may be used or useful in future jobs does not create a single dilemma at all for college students in how to actually complete their coursework. This is all nonsense & excuses for not doing the reading & learning how to think & write.
With all of the talks of AI Agents, I decided to outline what needs to happen for AI Agents to be successful.

And I condensed it into this silly diagram.

If you want to learn more read my latest article: www.chaos-engineering.dev/p/the-end-of...
Reposted by Francisco Javier Arceo
The rise of foundation models has incentivized a diversification of skillsets in both software engineering and AI research itself.

For most software engineers, being well rounded is more important than pure technical mastery. 1/3
Reposted by Francisco Javier Arceo
For anyone interested in fine-tuning or aligning LLMs, I’m running this free and open course called smol course. It’s not a big deal, it’s just smol.

🧵>>
Reposted by Francisco Javier Arceo
A special LLM course focusing on Smol models, the <2B parameters generic models which can run on-device/in-browser
For anyone interested in fine-tuning or aligning LLMs, I’m running this free and open course called smol course. It’s not a big deal, it’s just smol.

🧵>>
What's a feature in the context of AI/ML?

Anything that is input to a matrix used to generate a prediction.

That means:
- prompts
- context
- numbers in a matrix

And all of that can be used by a Feature Store
@bsky.app who do I have to beg to get code formatting in posts? 👀
prompt engineering = vibes engineering
Reject Vibes based AI!
anytime someone asks me how they should get started developing with LLMs
Evals are all you need.
anytime someone asks me how they should get started developing with LLMs
Reposted by Francisco Javier Arceo
anytime someone asks me how they should get started developing with LLMs
Reposted by Francisco Javier Arceo
Merkle Trees are such a powerful data structure for distributed systems. They can allow you save a lot of bandwidth for the small cost of re-computing hashes of data. They also are what allow you to trust data retrieved from untrusted sources www.baeldung.com/cs/merkle-tr...
How Do Merkle Trees Work? | Baeldung on Computer Science
A quick and practical guide to Merkle trees.
www.baeldung.com
The best part about working on open source is that you really do get to code for fun and for something that’s useful.
Reposted by Francisco Javier Arceo
Feature stores are the right foundation for serving data used for LLMs and GenAI.

A lot of different aspects of production AIbenefit from it.

It's what my entire goal is with Feast. I wrote more about in this article.

www.chaos-engineering.dev/p/production...

#machinelearning #ai #tech
Production Artificial Intelligence
On the future of Software and AI
www.chaos-engineering.dev
Who is building with Milvus? Making some progress for plugging it into Feast so that embedding and features can join the party.
I am going to be running to be a Kubeflow Steering Committee member and am very excited about the opportunity to make an impact.
I stopped using conda a while ago. So happy to see this result.