Thomas J L Mustard Ph.D.

I'm a computational chemist who became a product manager, and spent fifteen years doing it at the frontier of molecular simulation and AI. I grew up in the Pacific Northwest, studied chemistry at Eastern Washington University, and did my Ph.D. at Oregon State applying density functional theory to transition-metal catalysis: understanding how rhodium, copper, and palladium catalysts work at the electronic level, with results published in JACS, ACS Catalysis, and Angewandte Chemie.

During grad school I built Eta_Scripts, an open-source automation framework for transition state searching and reaction coordinate mapping. That project seeded an idea that has shaped everything since: the most leveraged thing you can do isn't to run the simulations yourself, it's to build the tools that let everyone else run them.

I spent nearly a decade at Schrödinger as Technical Product Manager and Principal Scientist, where I built the catalysis market vertical from scratch and launched AutoRW, an automated catalyst screening workflow that democratized a process previously requiring rare specialist expertise. Before AutoRW, fewer than 100 people worldwide could properly execute computational catalyst screening manually; AutoRW enabled 2,000+ screenings per year at 12× the cost efficiency of experimental synthesis, and secured deployments with Fortune 50 clients. I also drove product strategy across the advanced materials portfolio — polymers, semiconductors, organic electronics — including a key R&D collaboration with Panasonic, and supported pharma and biotech clients through the full DMTA cycle in LiveDesign, translating computational capabilities into accelerated small-molecule drug discovery programs.

Now I'm Staff Technical Product Manager at SandboxAQ, defining the product vision for an agentic AI platform that transforms the company's proprietary Large Quantitative Models (LQMs) and specialized datasets into standardized, enterprise-grade API products — the scientific Building Blocks that customers deploy to solve real physical problems in drug and materials discovery. The work spans designing Model Context Protocol (MCP) server roadmaps that position specialized models as domain-specific tools foundational LLMs call dynamically, leading product strategy for drug and materials discovery, and building multi-agent workflows that let medicinal chemists execute complex computational pipelines through natural language. Along the way I shipped an AI-powered invention disclosure workflow that compressed a two-week manual process into a sub-day automated pipeline, and hold two patents pending in AI agent systems, with a third filing imminent.

The through-line is the same problem at every scale: take something hard, expensive, and requiring rare expertise, and build software that puts it in more hands. The science is the entry point; the product is always the goal.