Member of Technical Staff, Machine Learning

About Mandolin

Nearly every disease will become treatable in our lifetimes. Mandolin is laying the clinical and financial infrastructure to get groundbreaking treatments to patients faster, powered by AI agents.

Mandolin partners closely with the largest healthcare institutions in the US, covering more than $10B drug spend across the country. We're backed by Greylock, SV Angel, Maverick, Signal Fire, and the founders of Vercel, Decagon, and Yahoo.

Our copilots handle ~80 % of clinic back-office workload today. Reaching 99 % means:

• capturing the long-tail edge cases that bury teams in rework,

• pushing the latest foundation models to their accuracy and cost limits, and

• proving every improvement with airtight, regulator-ready evaluation.

We need an applied-ML architect who has already built this bridge from “impressive demo” to “lights‑out production.” You will design the core ML infrastructure that turns Mandolin from helpful copilot into true autopilot—where work closes itself and clinicians only handle the exceptions.

• Full ML lifecycle. Architect HIPAA-compliant pipelines—data capture → training → inference → human-in-the-loop feedback to hardening and rollout.

• Model strategy. Select or fine‑tune SOTA vision, language, and multimodal models; apply LoRA, RAG, distillation, or quantization when ROI is clear.

• Evaluation harnesses. Build automated, statistical tests that surface long-tail failure modes and quantify business impact before any deploy.

• Performance engineering. Drive latency, throughput, and dollar-per-task down across GPU, CPU, and serverless footprints.

• Product integration. Pair with forward‑deployed engineers to turn field discoveries into new datasets, metrics, and rapid model iterations.

• Technical leadership. Mentor a junior ML engineer and establish best‑in‑class tooling, conventions, and review culture for the ML function.

• 5 + years shipping ML systems (vision, NLP, or multimodal) that serve live traffic.

• Expert Python plus PyTorch or JAX; comfortable with distributed training and experiment tracking.

• Track record of taking a research spike to a monitored, fault‑tolerant, low‑latency service.

• Deep understanding of eval design, ablations, and statistical significance in noisy real‑world data.

• Fluency with modern foundation‑model techniques:
  LoRA/PEFT, RAG, vector search, structured prompts, model compression.

• Bias toward measurable impact and disciplined rollout in regulated or high‑reliability domains.

• Healthcare, document understanding, or complex form‑extraction experience.

• Production work on Vertex AI, Kubeflow, or similar cloud ML stacks.

• Open‑source contributions, peer‑reviewed research, or public technical writing.