Filtration made green and easy: We're excited to share this News & Views article in Nature Sustainability, published with my graduate student Talia Khan. Whether on a hike, in a remote disaster zone or at home, access to clean water is critical. Filtration of freshwater to

We're excited to announce the 2024 Predictive Multiscale Materials Design short course: June 3-7, 2024, MIT Campus. Spend a week at MIT, earn a certificate, an learn about cutting edge modeling, design and atom-by-atom manufacturing methods for advanced materials. Generative AI

We can accelerate scientific discovery by teaching generative AI to make de novo predictions about never-before-seen ideas, concepts and designs. The method is based on a new approach that integrates: 1) Generative Knowledge Extraction, 2) Graph-Based Representation, and 3)

If you are interested in my MIT Professional Ed Predictive Multiscale Materials Design short course, please join our virtual open house on Thursday, April 4 at noon Eastern to get the inside scoop on the key skills you will learn, topics covered, and how the course will unfold.

Congratulations to everyone in LAMM@MIT and kudos to first author Rachel Luu for her excellent leadership in putting this together!

Thank you LlamaIndex 🦙 for highlighting our work!

Collagen is one of the most fascinating biomaterials - key to tendon, bone, skin, blood vessels and much more. I'll be presenting at the Collagen Café on 4/25/24 hosted by Andrew Rutenberg & Laurent Kreplak Dalhousie University. My talk will cover “Artificial Intelligence Driven Modeling

Check out mistral.​rs, our #Rust-based open source inference engine allowing for fast #LLM serving for a variety of architectures including X-LoRA mixture-of-expert (MoE) models, Llama-3, Mistral/Mixtral, Gemma & many others. Built on the Hugging Face #Candle framework for #Rust

Excellent work by our former grad student Eesha Khare!

Advanced Materials through Predictive Multiscale Design: What are the opportunities, challenges, and what does it take to advance the field for industrial impact? linkedin.com/pulse/advanced…

Thanks LlamaIndex 🦙 for highlighting our work! We recently added a couple of new features that you can immediately take advantage of together with the powerful LlamaIndex framework: ✅ In-situ 2, 3, 4, 5, 6 and 8 bit quantization: Create quantized models directly from

Vision Mixture-of-Experts (V-MoE) models allow us to integrate capabilities of several smaller models into a larger architecture: As part of our Cephalo series of vision-focused multi-modal V-LLMs for materials applications, we provide step-by-step examples of how to build Vision

The musicalization🎶of complex structures🕸, such as those found in biological materials ⛓ 🐜 🐚 🦋 , has been a longstanding focus of our research. Over the past decades, we have developed methods to transform the intricacies of material mechanics into musical compositions, and

How can we enable AI models to solve complex scientific tasks, in an autonomous self-improving manner? Agentic modeling is key! Multi-agent, multimodal AI frameworks allow us to integrate diverse types of data, like images, text, graphs, protein sequences & more. By understanding

Check out these cookbooks and open source codes that show how to build Vision Mixture-of-Experts (V-MoE) models "from scratch": We provide step-by-step examples of how to build Vision Mixture-of-Experts (V-MoE) and merged models, all in the Hugging Face ecosystem. Blog post:

Multi-agent, multimodal AI frameworks are enabling autonomous, self-improving models that seamlessly integrate with scientific simulation or experimental platforms like self-driving labs. Recent research shows how these agentic models tackle complex tasks such as protein

The challenge in #materials science is that are millions of possible combinations that could lead to the discovery of the next greatest, strongest, lightest, or most durable material. With the assistance of various #AI-powered tools, engineers can run through these combinations

This new paper from Massachusetts Institute of Technology (MIT) uses multiple #AutoGen agents to generate, refine, and validate research hypothesis! SciAgents: Automating Scientific Discovery Through Multi-Agent Intelligent Graph Reasoning by Ghafarollahi & Markus J. Buehler "We present SciAgents, an approach that

Thanks for highlighting our work AutoGen!

A fundamental challenge in #AI is the creation of systems capable of autonomously advancing scientific understanding by exploring novel domains, identifying complex patterns, and uncovering previously unseen connections in vast scientific data. We present SciAgents, a novel