Luis Capelo
@luiscape
Followers
2K
Following
2K
Media
77
Statuses
1K
Interested in deep learning research and applied AI systems. Building serverless at @modal
New York, NY
Joined May 2009
Real-time voice AI isn’t easy. It needs sub-second latency, natural turn-taking, and conversational quality all at once. Here’s how @DecagonAI and Modal built a real-time inference system using: • Supervised fine-tuning and reinforcement learning • Speculative decoding with
1
2
15
Only tech company that actually makes any sense right now is Modal.
5
3
85
Build a conversational voice bot with 1 second voice-to-voice latency with Modal, @pipecat_ai, and open models. Modal works seamlessly with WebRTC, WebSockets, and tunneling to squash latency to an absolute minimum.
8
23
252
smolagents is a popular library from @huggingface for building code agents. We’ve integrated with them so you can natively use Modal Sandboxes for secure code execution!
With @huggingface's smolagent v1.22.0 release, you can now use @modal Sandboxes for secure code execution. Just set `executor_type="modal"`! ☺️
1
3
30
The Modal SDKs for JavaScript and Go are now in beta, with support for creating Sandboxes and invoking Functions using Modal features like Images, Secrets, Volumes, and more.
2
4
36
The ACHIRA team is cracked. And now, for the first time, they have public openings for: - ML Researchers - GPU "make go brrs" - software engineers Build foundation models that simulate reality's most atomic units:
jobs.ashbyhq.com
Achira Jobs
We're launching ACHIRA, a newco DIMENSION cofounded alongside @jchodera & @tkaraletsos. We're building foundation simulation models for the smallest, atomic-level resolution units of the universe. Assembling here: https://t.co/LOoWMTe4UJ
0
5
36
We pushed Wan2.2 I2V inference to the edge, now 2.5x faster. Optimized with FA3 + TF32 + MagCache + Sequence Parallelism + torch.compile() — negligible loss in quality. Here's how we pulled it off:
morphic.com
Authored by Muhammad Ali Afridi. Introduction We have seen the rapid development of open-source Video Generation DiT models with MOE architectures, such as Wan2.1[1] and Wan2.2[2]. It is very...
3
18
120
You have to try https://t.co/Ho09ulo3vQ to really experience how fast this is. (Also, the playground is running on @modal 🙂)
playground.cognition.ai
Check out the Fast Context agent
Introducing SWE-grep and SWE-grep-mini: Cognition’s model family for fast agentic search at >2,800 TPS. Surface the right files to your coding agent 20x faster. Now rolling out gradually to Windsurf users via the Fast Context subagent – or try it in our new playground!
4
4
92
Here is quick tutorial on the quickest way to repoduce/run NanoChat with minimal GPU setup time using @modal
Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,
9
29
245
Write deep learning code on your laptop and run it instantly on GPUs - that's every developer/reseachers dream @thinkymachines offers this with Thinker You can do something similar by just using @modal right now !! I’ve written a blog + hands-on tutorial breaking down how you
13
72
704
We spoke with Modal CEO @bernhardsson about the "broken model" of massive GPU reservations. He says "hype" led companies to waste money on underutilized hardware. Modal's platform ensures you only pay for the GPU time you actually run.
3
11
93
our most cutting edge companies are consistently @modal power users now, @_DimensionCap is also a proud Modal investor. thrilled to join @bernhardsson & @akshat_b on their quest to reimagine infrastructure in the age of AI!!
It's true – @modal has raised a $87M Series B at a $1.1B valuation to advance the future of AI infrastructure. Thank you to @Lux_Capital, @Redpoint, @AmplifyPartners, and others. Now more than ever, AI demands a complete reinvention of traditional compute infrastructure
1
6
27
In case you missed @bernhardsson talking about Modal on @technology this morning!
Tune in at 8:45am PST/ 11:45am EST to watch @bernhardsson talk about our Series B announcement, and the future of AI Infrastructure on Bloomberg TV: https://t.co/VZgOrMJ8qa
2
7
66
🅱️ is for beating proprietary APIs 🅱️ is for Blackwell GPUs 🅱️ is for bus ads and, yes, 🅱️ is for billions
It's true – @modal has raised a $87M Series B at a $1.1B valuation to advance the future of AI infrastructure. Thank you to @Lux_Capital, @Redpoint, @AmplifyPartners, and others. Now more than ever, AI demands a complete reinvention of traditional compute infrastructure
8
6
164
the last 3 years have been the most fun building @modal. it truly feels like this is barely the beginning for us.
It's true – @modal has raised a $87M Series B at a $1.1B valuation to advance the future of AI infrastructure. Thank you to @Lux_Capital, @Redpoint, @AmplifyPartners, and others. Now more than ever, AI demands a complete reinvention of traditional compute infrastructure
0
0
11
Imagine we get to work with the cleverest people at @modal. Taking inspiration from FA4, we’re building optimized systems at many layers of the stack. Come work with us.
0
0
8
This is just a small part of CWM, with much more in the tech report! Truly amazing and my privilege to work with the incredible Meta FAIR CodeGen team😀. And thanks to @modal for the great sandbox support. Tech report: https://t.co/weJr89q6D4 Inference code:
0
6
24
We built next-edit autocomplete for JetBrains. It runs in sub-100ms with full codebase context:
20
17
252
I wrote an eng blog post: what's inside Notebooks Sharing how Modal boots real-time, collaborative GPU notebooks really fast. Full stack: - gVisor-based runtime (sandbox, filesystems, lazy image loading) - Multiplayer editing - Interactive streaming UI https://t.co/ryj2xt225a
6
28
283
Wow, thanks to @charles_irl , you can understand internals of vLLM with a live notebook from @modal 🥰
I had already planned to spend the day reading @gordic_aleksa's "Inside vLLM" blog post. That turned out to be an incredible fit for @modal Notebooks, released today! https://t.co/QKX1g9smdp
3
32
336