Thomas Pierrot

Our ChatNT work just made the cover of the June issue of @NatureMachIntell 🤯 "A conversational agent fluent in biological sequences" So incredibly proud of the dream team at @InstaDeepAI! What a milestone ⭐

Introducing ChatNT: The first biological sequence-language model, published in @NatMachIntell ! 🧬🎉 Inspired by vision-language models, ChatNT's architecture combines biological and language foundation models, using Nucleotide Transformer (NT) and Llama to answer questions…

Excited to share our latest work applying Bayesian Flow Networks (BFNs) to proteomics! We show how BFNs can outperform leading autoregressive, discrete diffusion, and BERT models in protein sequence modeling. 🧵

holy shit AI can now decode the human proteome without any reference genome. a new paper published today presents a new model that can read raw mass spec and generate novel peptide sequences from scratch. no database or priors. just biology. 1/

🤖 The action at the #AIActionSummit continues tomorrow with "Foundations and Advances in Generative AI" organized by the #ML Department of @mbzuai. 💬 Attendees, don't miss out on @thomas_pierrot's session on Multi-modal Foundation Models for Biology, 13 February at 10:10 AM .

Looking forward to it!

As the #AIActionSummit kicks off in Paris today, our France Lab is excited to announce its first workshop "Foundations and Advances in Generative AI". 📅 When: February 12–13, 2025 📍 Where: Fondation François Sommer, 62 Rue des Archives, 75003 Paris, France Co-hosted by our…

Recurring question: So what? It can't be used to train Deep NNs, let alone LLMs. That's true. ES doesn't work well in high dimensional spaces. As it maintains a population, memory requirements (if you are using ES to train Deep NNs) are prohibitive and sample efficiency is not…

From breakthroughs to product launches and more, it has been a stellar year at @instadeepai! ✨ Wishing you all happy holidays 🎉 AI is the Force, let's steer it to benefit everyone in 2025! 🌍🌟🚀

BulkRNABert: Cancer prognosis from bulk RNA-seq based language models 1. BulkRNABert is the first transformer-based language model pre-trained on bulk RNA-seq data, designed for cancer type classification and survival analysis. It leverages self-supervised learning to create…

Learning the Language of Protein Structure • A novel approach to protein structure modeling: this study introduces a vector-quantized autoencoder (VQ-AE) to transform continuous protein structures into discrete representations, enabling the use of natural language processing…

On my way to @NeurIPSConf!✈️ Come chat with the @instadeepai team about our latest work on foundation models for biology. Presenting exciting new research! Also keen to connect with folks using the Nucleotide Transformer and anyone interested in ML for science. 🧬 See you there!

🎉 Big news! Our Nucleotide Transformer foundation models for genomics were just published in @naturemethods! 🚀 So proud of this incredible team @instadeepai! ⭐Paper: go.nature.com/3OA7dWr 📕Research briefing: go.nature.com/3BbSPQY

Our Kyber supercomputer makes it to the top 20 H100 GPU compute index! 🔥🚀

Nice to see this launch of a Pro version of DeepPCB, an RL-based system for laying out PCB boards, which runs on @googlecloud. Nice work, @kbeguir and team! 🎉 The use of RL for many kinds of hardware design and layout problems is accelerating. I remember excitedly comparing…

👋 Hello World! 🌎 Introducing DeepPCB Pro, our AI-powered design tool for industry-grade PCBs! Fast, precise, and compute-scalable with @googlecloud ☁️ Check it out at deeppcb.ai 👀