Broad by instinct, deep by choice.

Context

Research assistant on projects involving large language models, graph unlearning, and topological graph structures. Primary responsibilities: replicating results from existing papers, implementing experimental pipelines, and supporting ongoing research. Second author on a paper exploring the use of LLMs for node and graph unlearning tasks.

Constraints: limited compute budget, tight publication timelines, no dedicated infrastructure.

Action

• Inference 4x faster — profiled HF Transformers serving path; identified KV-cache thrash and tokenizer overhead as bottlenecks.
• Evaluated vLLM — concluded setup cost outweighed gains under research timeline; documented the call.
• Custom data pipeline — assembled a one-off ingestion stack for unstructured experiment logs using third-party tools.

Lessons

• Sophistication has overhead. A simpler CUDA setup outperformed a more advanced serving framework given the constraints. Knowing when not to use a tool is as important as knowing how.
• Economic thinking applies to research. Time spent building infrastructure is time not spent running experiments. I learned to evaluate build-vs-buy tradeoffs quickly.

Context

Analyzed Phase II clinical trial data for an MDD drug at a pre-IPO biotech company. Objective: use ML and statistical methods to identify potential gaps, latent signals, or patterns not captured by traditional analyses. I had no prior experience in clinical trials or biostatistics.

Constraints: small trial size, limited data points, high noise, and significant placebo effects—conditions that are common in psychiatric trials and fundamentally limit what ML can detect.

Action

• Trained classification models on biomarker data; XGBoost performed best of the candidates.
• Honest negative result — accuracy unstable across folds; recommended against production deployment.
• Domain ramp-up in 2 weeks — paired with biologists to translate assay outputs into ML-ready features.

Lessons

• Knowing when ML doesn't apply is part of the job. Small, noisy clinical datasets often lack the statistical power for ML to add value. Recognizing this early prevents wasted effort and misleading conclusions.
• EDA and domain understanding matter more than model complexity. The most useful insights came from careful exploratory analysis, not from algorithmic sophistication.

Context

GitRoll is a pre-seed HR-tech startup that scans GitHub repositories to evaluate developer contributions—designed to surface signal beyond resumes and help underrepresented developers get fairer evaluations.

Action

• Scaled to 45k+ users — owned PR-classification pipeline end-to-end; ~92% precision on 216+ hand-labeled issues.
• Cut infra cost 13% — benchmarked GPT, Gemini, Claude; optimized model selection and prompt compression.
• +20% throughput — replaced word-bank heuristic with NLTK-based categorization after A/B comparison.

Lessons

• Opportunities can be created, not just found. Proposing a trial period was uncomfortable but effective. Execution speaks louder than credentials.
• Early-stage startups reward breadth. I touched LLM systems, labeling, static analysis, and product iteration—all within a few months. Context-switching is the job.