Jacob Phillips

Jacob Phillips is currently building Applied Compute, a company focused on 'Specific Intelligence' which involves creating custom models and purpose-built agents for enterprises. The company has successfully raised $80M in funding and is actively hiring across research, infrastructure, and engineering. nitter

He is an Engineering Fellow on Andreessen Horowitz's American Dynamism team, where he focuses on AI and robotics. In this role, he has been involved in open-sourcing engineering and research projects, notably ARES, and has written extensively on topics like 'How to Build ChatGPT for Robotics' and 'ARES: An Open Source Platform for Robot Data'. a16z+2

Jacob has a strong background in Machine Learning and AI research, having worked as a Machine Learning Research Engineer at Scale AI where he trained their Automotive Foundation Model and built LLM agents. He also co-founded Themis AI, a startup that focused on delivering debiasing solutions for machine learning algorithms. equilar+1

He is a graduate of MIT, holding both a B.S. and an M.Eng. in Computer Science. His academic research includes his Master's thesis on 'Unsupervised Latent Debiasing of Time-Series Models', reflecting his interest in fairness and bias in AI, computer vision for autonomous robotics, and AGI alignment. jacobdphillips+1

Jacob is actively engaged in developing and open-sourcing tools and benchmarks for AI and robotics. His notable projects include ARES, a robot data platform; MALLET, a multimodal toolkit for VLM evaluation in robotics; DailyBench, a benchmark for detecting model regression in frontier LLMs; and OpenRubric, a framework for multi-step rubric evaluation. github+2