Ryan Kim

DNABERT-based prognostic model for glioblastoma survival via Cox regression on non-coding regulatory DNA mutations; identified age-related survival signals and racial disparities encoded in mutational scores.

Raised $20k in funding (Hudson River Trading, Innovate901 First Place). Built a full-stack math competition platform serving 150k+ members across 11 partner organizations.

Engineered a CPU-only, locally-deployed LLM clinical pipeline (LLaMA, Qwen, DSPy) to classify lung cancer CT reports across 370 cases; co-authored a referral-event analysis abstract accepted at the ACRO 2026 Summit. Applied Cox regression and survival analysis on 230 patients, linking percent thymic tissue to NSCLC outcomes.

Benchmarked 12 large pathology foundation models (e.g., UNIv2, Prov-GigaPath) across TCGA and internal cohorts, quantifying a 2x generalization gap in AI tumor-purity prediction and revealing clinical deployment risks.

Founded platform indexing 500k+ ML/AI papers lacking public code implementations; built an automated paper-to-GitHub matching pipeline that surfaces which papers still have no public implementation.