We develop scalable and self-improving AI systems and methodologies towards the goal of AGI, leveraging techniques in machine learning, systems, natural language processing, and beyond.

Prospective Students and Postdocs:

Research Themes

Scalable Intelligence

We build new AI systems that scale robustly across data, parameters, and compute.

Agentic Optimization and Self-Improvement

We create agentic frameworks that enable machines to learn from their worldly interactions and experiences, enabling advanced reasoning and planning.

Automated Verifiers and Benchmarks

We design methods and frameworks for automated evaluation and benchmarking.

AI for Systems and Systems for AI

We design efficient systems for AI training and serving, with a particular interest in using AI itself as the optimization engine.

Courses

CS329A - Self-Improving AI Agents

This course covers the latest techniques and applications of AI agents that can continuously improve themselves through interaction with themselves and the environment.

CS229S - Systems for Machine Learning

This course focuses on performance efficiency and scalability of deep learning systems, covering efficient training, fine-tuning, and inference with an emphasis on Transformer architectures and LLMs.

People

Masters & Undergraduate Researchers

Alumni

Design from MIT Visualization Group