Last fall, we shared our deep dive on FA4 internals.
But we didn't stop at grokking the kernel.
Since then, we've been developing improvements for inference performance and upstreaming them.
This blog post explains those contributions.
https://t.co/xzDNHdq3Zw
At @modal, we're working to make sure OSS RL frameworks have all the techniques necessary to train frontier open-weights models.
Delta compression is key, but the job's not done. There are still lots of open problems around weight sync, auto-scaling, & cross-cluster training.
My DMs are open!
We're honored to deepen our partnership with @modal and co-lead their $355M Series C!
When @bernhardsson and @akshat_b started Modal in 2021, they had conviction that building a truly great cloud for AI meant rebuilding the entire stack from the ground up. Five years later, that bet has paid off.
Erik built the recommendation system behind Discover Weekly at Spotify. Akshat was an early engineer at Scale AI before joining Erik as co-founder and CTO. Together, they've assembled what one engineer affectionately calls "a monastery for super nerds" and it shows in the product.
Modal has become the high-performance cloud that serious AI teams reach for when they need to ship. We couldn't be more excited to continue backing them.
Huge congratulations to Erik, Akshat, and the entire Modal team!
Raising $ is cool. What’s even cooler is getting to work every day with this incredible group of humans.
We like solving hard problems and building things we can be proud of. If this is you, come join us! We’re just getting started :)
Today we're announcing our Series C funding: $355M at a $4.65B valuation, led by some great investors @generalcatalyst and @Redpoint.
We've had insane growth in the last year, but we're still very early. So proud of the team and what we have built so far!
"Sandboxes are one of the most important building blocks for reinforcement learning."
- @ypatil125 on why every rollout needs its own clean, replayable environment, and how that single requirement shapes everything else in the training loop.