Remote: Yes (also open to onsite in Boulder County, CO)
Willing to relocate: No
Technologies: Python, SQL, applied ML, RAG/LLM systems, embeddings, evaluation/guardrails, signal processing, data pipelines, instrumentation, optics/photonics, C++, MATLAB, AWS
Résumé/CV: https://robheffern.palaich.net/cv/
Website: https://robheffern.palaich.net
Email: Robert.Heffern@gmail.com
Applied physicist and engineer, 15+ years. I build ML and LLM systems that turn messy, high-stakes data into something a human can trust and act on; and I want it aimed at patients. Put me on a problem where being wrong has consequences and the thing has to ship, not demo.
Recent: 6 years at Agilent building production ML over instrument telemetry: fleet-wide failure prediction, and a RAG/LLM system for 300+ field engineers taken from prototype to production. I owned the confidence logic for when it answers vs. hands off to a human, the part that matters when being wrong has cost. Cut data-prep time 8x.
I know data from the sensor up, not just after it lands in an S3 bucket: before Agilent I built scientific instruments from scratch (a fiber-based STED nanoscope and an OCT microscope). Earlier: DoD hardware reliability (TS/SCI), precision timing, published detector-physics work.
Best at the seam where software meets the physical world and the answer has to be right: sensor data, ML you can trust, and knowing (and saying) when a result isn't trustworthy. On my own time I build systems that turn large structured knowledge bases into narrative a person actually wants to read; making complicated things legible is the throughline (see website).
Open to full-time or contract. I'm perfect for for health / medical-device / clinical-AI missions.