Research Lead at Prime Intellect, building decentralized compute infrastructure for distributed AI training. Creator of the verifiers toolkit for RL environments. Previously on Morgan Stanley’s ML Research team, bringing deep expertise in reinforcement learning and multi-agent systems.
Pioneer in Democratized AI Research
Will Brown champions radical accessibility in AI research, arguing that talent—not compute—is the key constraint. His mission centers on breaking open the “black box” of model training through open-source infrastructure and community-driven development.
Current Work
As Research Lead at Prime Intellect, Will spearheads infrastructure for decentralized AI training at scale. His work focuses on the verifiers toolkit, an open-source Python library for building RL environments and training LLM agents. He emphasizes that “environments are the webapps of research”—fundamental abstractions that define task specifications, evaluation frameworks, and reward signals.
Key projects include the Environments Hub for creating and sharing RL training environments, distributed compute infrastructure enabling training across decentralized GPU networks, and research on multi-turn RL with turn-level credit assignment for multi-hour agent tasks.
Blog: Five years of feedback loops - PhD summary and future directions Blog: ParaLLM: 1600+ tok/s on a MacBook - Fast parallel LLM inference Blog: Generative AI Handbook - Curated learning resources Podcast: Latent Space - Multi-Turn RL for Multi-Hour Agents
Background
PhD work centered on multi-agent learning and reinforcement learning systems at Columbia University. Previously ML engineer at Morgan Stanley’s Machine Learning Research group, developing production ML infrastructure.
Philosophy on Research
Will’s approach challenges elite research gatekeeping:
Research for collective understanding - Advocates for open research that advances shared knowledge rather than proprietary capabilities, with the contrarian stance that talent scaling matters more than compute scaling.
Accessibility over exclusivity - Breaking down barriers to serious AI research by providing simple, powerful tools that individual researchers can use effectively.
Environments as key abstraction - RL environments encapsulate everything needed for reproducible research: task definition, evaluation harness, and reward signals.
Prime Intellect
Prime Intellect builds decentralized compute infrastructure for distributed AI training, making advanced model training accessible beyond elite research labs. The platform enables researchers to leverage distributed GPU networks for large-scale experiments through open-source tools and simple infrastructure.
Conference Appearance
Event: AI Engineering Code Summit 2025 Date: November 21, 2025 Time: 11:40 AM - 11:59 AM Session: RL Environments at Scale
Will presented on scaling reinforcement learning environments for production use, covering Prime Intellect’s decentralized compute platform, the Environments Hub for sharing RL setups, and practical patterns for distributed training. His talk emphasized making model training accessible to all researchers through open-source tools.