Research
In Action

Evaluating LLMs for Automated Measurement

LLM Evaluation · NLP · Research Engineering

Working alongside Brian Yim and Dr. J. Christopher Muran, I investigated whether large language models can reliably score psychotherapy transcripts using the Experiencing Scale (EXP). We benchmarked six models total: Claude, GPT-5.4, and Gemini in the cloud, alongside Llama 3.1, Gemma 3, and Mistral running locally on private Google Cloud GPU infrastructure. I engineered the privacy-preserving de-identification pipeline (Safe Harbor 18 identifiers via Microsoft Presidio and spaCy), solved enterprise zero-trust networking (IAP tunneling, Spot GPUs, Cloud NAT), built a fault-tolerant Python scoring pipeline executing 825 model runs with zero data loss through VM preemption, and analyzed the empirical factors (emotional peak sensitivity and score variance restriction) influencing inter-rater reliability.

Computational Origins of Biological Energy

AI × Computational Biology

Under Dr. Brian Stockman's mentorship, I'm investigating how the first electron transfer proteins emerged at the origin of life — exploring whether small peptides built from only the 10 prebiotic amino acids could fold into functional structures capable of transferring biological energy. In this research, I leverage AI, machine learning, and computer science using tools like AlphaFold, Rosetta (Baker Lab), MadraX, and custom Python code to model, simulate, and analyze prebiotic structural space.

Untitled Research in ML for now

Machine Learning · Artificial Intelligence

A new machine learning research project starting soon. Further details, theoretical framework, experimental benchmarks, and preliminary findings will be shown here as the investigation gets underway.