Overview

Lumos AI Research builds data infrastructure to improve AI model accuracy and safety for health and life sciences. It partners with labs, startups, and governments and has supported major tech firms like OpenAI, Google, and Meta. Its offerings center on data-centric AI methods including Supervised Fine-Tuning (SFT) with high-quality prompt-response examples, Reinforcement Learning with Human Feedback (RL/HF) guided by medical experts, Preference Ranking, and Reinforcement Learning from AI Feedback (RLAIF). It also develops Health RL simulation loops and uses Traces to teach AI agents from user actions. The company differentiates itself through a focus on high-stakes health data, a team with backgrounds from OpenAI, Scale, TikTok, MIT, and Harvard, and its role as a partner for training and evaluating models in regulated environments. Its goal is to help organizations build, train, and evaluate safer, more accurate AI systems for health and life sciences.