I'm the cofounder and CTO of Photonium, where we build AI tools for optical design. Before that I was an undergraduate physics researcher at Stanford, working on the spatial clustering of active galactic nuclei, and a computational physicist at a fusion startup.

Work

Photonium, cofounder & CTO
  • Built and maintain a production agentic AI system for optical engineering (cameras, LiDAR, semiconductor lithography, medical imaging), enabling engineers to query, analyze, and iterate on complex lens systems through natural language.
  • Developed a native parser for industry-standard optical design files (.zmx), enabling structured retrieval and reasoning over proprietary design libraries without existing open-source tooling.
  • Designed a robust evaluation framework benchmarking performance across aberration diagnosis, optical prescription generation, and design execution stages.
  • Built observability and telemetry layer to detect and surface agent failures in production, enabling rapid prompt iteration and regression tracking.
Avalanche Energy, computational physicist
  • Simulated fusion-relevant plasmas using WarpX, a Particle-in-Cell code, to explore various parameters and their effect on the fusion energy gain factor (Q value) value for the Orbitron, an electrostatic fusion reactor.
  • Work included running simulations on AWS GPU clusters, writing analysis scripts, using Paraview (a scientific visualization tool), presenting and summarizing my findings in a detailed write-up.
Stanford XOC / KIPAC, undergraduate researcher
  • First-author publication: “AGN with Massive Black Holes Have Closer Galactic Neighbors,” Astronomy & Astrophysics (2025).
  • Pioneered use of kth-Nearest-Neighbors in the context of measuring Active Galactic Nuclei (AGN) clustering, discovering more massive AGN tend to rely in denser cosmic environments with a confidence level of 4.8 SD

Writing

All writing →

Background

  • B.S. Physics (incomplete), Stanford. Left spring of junior year to start Photonium.

Elsewhere

© Adam Mhatre, 2026Last updated April 2026