Founding Engineer in San Francisco · Physics-informed ML & simulation, from research to systems
I build physics-grounded ML and multiscale simulation — from DFT to coarse-grained MD — and turn it into working systems. I recently joined a San Francisco startup as a founding engineer. Before that, a Georgia Tech Ph.D.: an ML closure for polymer theory (SANS-validated, >90% benchmark wins) and force-field tooling that cut binding-energy error 46%.
ACS PMSE Emerging Professional Award 2026 — 1 of 15 selected worldwide
A physics-informed machine-learning closure for polymer integral-equation (PRISM) theory — trained on 370+ GB of coarse-grained MD across ~400 systems, SANS-validated, and beating classical closures in >90% of tests, with inference in minutes instead of days.
Know more →pyForceTune — a Python/Pyomo toolkit that optimizes bonded and nonbonded force-field parameters against DFT reference data, cutting binding-energy error 46% with <0.1 kcal/mol dihedral error for reliable conjugated-polymer simulation.
Know more →Multiscale DFT/TDDFT with atomistic and coarse-grained MD, linking molecular design to self-assembly, morphology, and device-relevant behavior in organic electronics.
Know more →
Currently a founding engineer at an early-stage San Francisco startup (ML / simulation systems) — happy to compare notes with builders, founders, and people working on hard ML problems.
Best way to reach me: zfeng77@gatech.edu