MSc ETH Mechanical Engineering · Zürich, Switzerland Applied AI in regulated environments — healthcare and pharmacy. I build systems that process sensitive patient data, so almost all of my code is private by design (proprietary, covered by trade-secret agreements). This page is the map; the public work sample is rag-eval-harness.

LinkedIn · Google Scholar · nicholas.buenger@gmail.com

FarmaciaOS — pharmacy management platform (private) Modular ERP for Swiss pharmacies: dispensing, accounting, controlled-substances registry, audit trails, robotic dispensing (WWKS2), e-prescription, MediData/HCI e-health integrations. Ships an on-premise AI module: an open-weight LLM served over a RAG layer on the pharmacy's own data — all patient data stays on-site. Live in production. Node.js/Express · React · MySQL · Python · Ollama

RicDictate / Medivia — clinical voice-to-report (private) Radiology dictation pipeline: Whisper-family ASR → LLM structuring → signed PDF report, fully on-premise with de-identification. Fine-tuned ASR and language models; systematic evaluation against real cases with a plain-language feedback loop that corrects the system over time. Built side-by-side with a practising radiologist. Python · PyTorch · on-prem NVIDIA

farmaciadelparco.ch — omnichannel pharmacy e-commerce (private) Web shop + native iOS app for my own pharmacy: catalogue, payments, logistics, multilingual search, and an autonomous LLM agent that monitors and patches the production system. Designed, built and operated single-handedly. Vanilla JS · Node.js · MySQL · Cloudflare · Swift

Balgrist University Hospital — data & AI platform (institutional) Product Owner & Data Engineer of the Biomedical Informatics Platform shared by four Zürich university hospitals, ETH and UZH (>1M patients). Deep-learning on medical imaging (DIZH MedTwins, PyTorch), the hospital's LLM environment, and the only NVIDIA DGX A100 in a Swiss hospital — air-gapped. Co-built the OR-X national smart-OR platform (arXiv:2503.14049).

HospitHome — tele-monitoring for home care (2020–2022) Real-time streaming and processing of vital-signs sensor data; 700 patients monitored over two years as CTO. Co-author of the SUNSHINE-eHealth study.

• rag-eval-harness — reproducible harness for evaluating RAG pipelines (chunking, retrieval, answer quality) on a synthetic corpus. numpy-only, tested, CI. It shows the method behind my production RAG systems without exposing them.

• NVIDIA GTC 2024, San Jose — invited speaker (alongside Microsoft) · HIMSS Europe 2023, Lisbon — speaker
• Second author, "MedSAM2-Anatomy: Training-Free Inference-Time Optimization for Musculoskeletal Segmentation" — arXiv:2608.00195 (2026)
• Co-author, "Digital Twin Platform Integrated into a Hospital Trusted Research Environment for Hip Preservation Surgical Planning Research" — MICCAI 2026, DT4H Workshop (accepted)
• First author, IEEE/RSJ IROS 2021 (with Prof. Daniela Rus, MIT CSAIL)
• NVIDIA Certificate of Competency — "Deploying and Optimizing AI Inference at Scale" (2026)

Want to see the private systems? Happy to demo them in person.