AI Architect & Builder - Inference, LLMs, SLMs GSoC Org Admin - Jenkins | Docker Captain | CNCF Ambassador

New Delhi
Joined December 2018
What an amazing day! Presented a poster at @PyTorch Conference: how we optimised the RLM paper with pre-fix caching & batched Sub calls with @vllm_project And a full house for our Pytorch Conference talk on running LLMs using Executorch on Android!
4
5
2
89
13,665
Shivay Lamba retweeted
I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined. I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top. The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever. My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply. To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had. For me, it is time for a bit of a break before diving into something new. If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.
1,890
875
704
25,474
3,243,907
Cached tokens are now live on Nebius Token Factory⚡ Reusing context across requests? Cached input can help reduce costs. See cached-token usage in API responses and billing, with pricing per model. More info: docs.tokenfactory.nebius.com…
5
17
13
125
22,916
Shivay Lamba retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model. 🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal. 🔹 Full-project coding: games in Unreal, Unity & Godot. 🔹 Excels at parametric 3D modeling in Blender & CAD. 🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%. 🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks. Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Open weights by Nov 4, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
1,139
2,883
2,188
26,672
2,940,023
Shivay Lamba retweeted
Meet Kimi K3.1 Pro, our most capable 3T-class model.
414
1,402
396
24,344
1,588,044
past ~48 hours in security: > PHP: unauthenticated remote code execution > GitLab: CVSS 10.0 remote code execution > Node.js: Brotli decoder RCE + 9 other flaws > Chrome: 353 security fixes > Signal: a bug that can prevent calls fully ending > eMudhra: compromised CA keys > Oracle: 1,295 security patches > Fortinet: critical/high fixes > Atlassian: 176 vulnerability entries and that’s after the Linux VM escape disclosed last week. ask Codex / your agent to check your code, dependencies and infrastructure for exposure to these. set up automations to regularly check for newly disclosed vulnerabilities and flag what needs patching.
39
62
18
712
48,615
Shivay Lamba retweeted
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
627
886
625
8,442
3,386,735
Shivay Lamba retweeted
A lot of people trying to imply that last 1 year of vibe coding is responsible for all the security vulnerabilities dropping across every piece of software this week. But if you take a deeper look you’ll see most of the CVEs are impacting actually hand written code from the past; and in fact AI agents are super important to defend against cyber attacks (quickly deploying changes at scale, monitoring situations, writing the code to patch the exploits etc). Just because humans wrote it doesn’t mean it is necessarily more secure than what AI wrote.
38
18
4
367
19,196
We’ve spent the past year developing new primitives for GPUs. We believe this enables new algorithms to change how machines learn, open new markets, and dramatically reshape AI’s cost structure. We have raised a $1 billion Series A, co-led by Index and Lux.
289
221
237
3,766
829,045
Shivay Lamba retweeted
Had a wonderful time presenting 2 talks at @AllThingsOpen: 1. How do you secure coding agents using @Docker Sandboxes 2. LLM inference on edge using @PyTorch Executorch and LiteRT for @Qualcomm_Dev QNN Always have wonderful discussions and cross sharing of feedback at ATO!
8
1
47
1,068
Had a wonderful time presenting 2 talks at @AllThingsOpen: 1. How do you secure coding agents using @Docker Sandboxes 2. LLM inference on edge using @PyTorch Executorch and LiteRT for @Qualcomm_Dev QNN Always have wonderful discussions and cross sharing of feedback at ATO!
8
1
47
1,068
Also great to catch-up with some of the best folks in the business after such a long time: @kentcdodds @ossia @fharper @SwiftAlphaOne This is only possible at ATO 🫡
2
8
129
Shivay Lamba retweeted
New research: We developed JOLT, a post-training method where one model acts as both teacher and student, with the teacher trained to give feedback the student can use. On MATH, it matched GRPO's accuracy with 6.5x fewer training tokens. perplexity.ai/hub/blog/a-goo…
68
55
18
592
82,046
Shivay Lamba retweeted
Today we release mammoth-1 and elephant-1! We post-trained mammoth-1, the #1 ranked best decision model in the world, and elephant-1, which is the best model under 15B (and beats Jev). We're open-sourcing all model weights and provide a technical report detailing all our training and data methods.
21
64
36
729
85,599
Shivay Lamba retweeted
We teamed up with @huggingface to give away $20,000 to anyone who's building with Beam 🤗 Use Beam for any project - make a demo or a dataset, benchmark the model, create an RL environment - and comment your link in the model's repo page by Monday, November 2. We choose the winners. Winners choose cash or credits. Claim $25 in Hugging Face credits to build with Beam and enter the competition 👇 huggingface.co/reflection/Be…
13
30
10
346
65,057
Shivay Lamba retweeted
Inspired by @BhosalePratim's Macbrow, I built Winbrow: a Windows voice assistant. @GradiumAI speech + @typesafeai Jev routing + browser automation adapted from @browser_use’s Jev Ultrafast project. @Alibaba_Qwen 4B LLM handles replies. Phonon (TTS) speaks them both running locally on @Qualcomm_Dev X Elite’s NPU. Built for Phonon’s OSS launch!
I've always wanted a way to talk to my Mac and do things on the go. Until now, it was super slow. So I built Macbrow. Voice-controlled browser and computer-use agent. Built using @GradiumAI @browser_use and @typesafeai Still early and a work in progress. github.com/timpratim/macbrow
6
3
34
1,369
Shivay Lamba retweeted
Model customization recipes are live on our website and in the cookbook repo. Beam is now supported across RL and SFT stacks including @axolotl_ai, NeMo RL, NeMo AutoModel, Megatron Bridge (@NVIDIAAI), Miles (@radixark), SkyRL (@NovaSkyAI), and TRL (@huggingface). Get cooking 🧑‍🍳 developers.reflection.ai/cus…
5
9
6
121
7,947
Inspired by @BhosalePratim's Macbrow, I built Winbrow: a Windows voice assistant. @GradiumAI speech + @typesafeai Jev routing + browser automation adapted from @browser_use’s Jev Ultrafast project. @Alibaba_Qwen 4B LLM handles replies. Phonon (TTS) speaks them both running locally on @Qualcomm_Dev X Elite’s NPU. Built for Phonon’s OSS launch!
I've always wanted a way to talk to my Mac and do things on the go. Until now, it was super slow. So I built Macbrow. Voice-controlled browser and computer-use agent. Built using @GradiumAI @browser_use and @typesafeai Still early and a work in progress. github.com/timpratim/macbrow
6
3
34
1,369
How it works: 🎙️ Microphone → Gradium STT 🧭 Jev → chooses the action 🖥️ Windows controls / browser agent → executes 🧠 Local Qwen → answers questions 🔊 Local Phonon → speaks replies STT and routing use cloud APIs. Qwen and Phonon inference run on-device.
2
169