Liam Kozma

I build and scale AI-enabled biological systems. My research focuses on the intersection of protein language models, high-dimensional statistics, and high-performance computing.

Research

Investigating Recovery Thresholds and Phase Transitions in Protein Language Models.

Protein Language Models · Phase Transitions

An analysis of mathematical phase transitions, specifically BBP phase transitions in spiked covariance models, and compute-optimal scaling laws within generative biological AI. This research focuses on how biochemical feature contributions cross critical thresholds to enhance sequence recovery in inverse folding tasks.

Data Diversity and Sequence Length: Key Levers for Powerful Biological AI.

Benchmarking · Out-of-Distribution Evaluation

Co-authored research introducing Genetic Stratification for Inference in Molecular Modeling (GRIMM). This work establishes rigorous benchmarks for out-of-distribution evaluation and sequence clustering, addressing the critical challenge of low sequence diversity in predictive enzyme function models.

Engineering

Applied Research Scientist - US Army Corps of Engineers.

Contract work focused on applied research and development within nanotechnology, delivering highly technical, scalable solutions tailored to defense-oriented scientific applications.

Details withheld: work performed under federal nondisclosure. No deep-dive available.

Bioprocess Optimization via Evolutionary Algorithms.

Engineered a stiff fed-batch bioreactor simulator for L-asparaginase production in metabolically engineered E. coli, then drove it with two gradient-free optimizers. A particle swarm (PySwarm) tunes the PID temperature controller; a genetic algorithm (DEAP) evolves the reactor design for maximum profit. The model is blunt about its own economics: the process pays only once acetate overflow is engineered out of the strain.

Off-Script

When I am not scaling models or analyzing biological data pipelines, I train MMA and Brazilian jiu-jitsu, and play Spanish guitar.