Nature Sensors co-author · research & software engineer · building toward ML & AI interpretability

I build systems that extract and reveal hidden structure in complex information. 2026 UW–Madison graduate triple-major (Mathematics · Physics · AMEP (Applied Math, Engineering, and Physics)).

Research — Co-author, "Robust Spectral Sensor for Standoff Biometric Detection," Nature Sensors (Jan 2026); implemented and benchmarked the classical baseline methods for the rPPG pipeline behind it (frequency-domain signal extraction + motion-compensated tracking).

What I build

• rPPG — the Nature-paper pipeline: heart rate from video via FFT/wavelets + Lucas–Kanade stabilization.
• Knowledge_Graph_Builder — LLM system (Anthropic API) building interconnected knowledge graphs from source data.
• Agentic LLM systems (provenance-tracked doc generation; a pipeline that generates working, tested repos from a spec) + a production Flask app (hardened auth, 91-test suite, deployed).

Now — focused on ML engineering & mechanistic interpretability; studying a topological-data-analysis view of grokking.

Python (NumPy, SciPy, OpenCV) · LLM/agentic pipelines · math (group theory, real analysis, linear algebra) · DSP (FFT, wavelets)

hschn4@gmail.com · LinkedIn · open to ML / Research Engineer roles