SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Novellia is a patient-powered real-world data platform that enables patients to access and control their health records, transforming complete health journeys into datasets for biopharma innovation and research. Founded in 2023, the company has raised ~$30M from tier-1 investors (Spark Capital, Khosla Ventures, Bling Capital) and is growing 5x year-over-year.
You'll be Novellia's first ML hire, joining the Platform Engineering team and reporting to the Head of Platform Engineering. Your core mission: transform unstructured clinical text (notes, discharge summaries, pathology reports, scanned documents) into trustworthy, structured features that make longitudinal health histories usable for research.
The role is split roughly 70% applied ML on clinical text (entity extraction, classification, sequence labeling, annotation strategy, error analysis, calibration, evaluation discipline) and 30% LLM-based work (prompt development, structured output, retrieval, evals, observability). You'll own the full lifecycle: framing problems, data/annotation strategy, model selection, training/fine-tuning, evaluation, deployment, monitoring, and retraining. This is not a research seat or a hand-off role—you set technical direction and make the interesting architectural decisions about what to extract first, how to validate correctness, and what a mature extraction pipeline looks like at scale.
Key responsibilities include defining "accurate enough" with clinical and customer-facing stakeholders, building defensible evaluation harnesses, partnering with Clinical Data Managers on curation and QA/QC, productionizing NLP pipelines against messy real-world data, applying LLM rigor (versioned prompts, real evals, cost/latency tracking, known failure modes), and making extraction quality legible to non-ML colleagues. You'll treat de-identification, PHI handling, audit trails, and access controls as part of the modeling problem.
Required: 6+ years applied ML with models taken from problem to production; healthcare or life sciences experience with real clinical data (clinical notes, EHR, claims, registries); depth in applied ML on text (NER, classification, sequence labeling, weak supervision, annotation guidelines, inter-annotator agreement); practical LLM experience (prompt development, structured output, retrieval, fine-tuning, evals); strong Python engineering fundamentals; strong cross-functional collaboration; self-directed problem-solving mindset.
Nice-to-have: clinical terminologies (SNOMED CT, ICD-10, LOINC, RxNorm, CPT, FHIR); HIPAA/SOC 2/de-identification/IRB experience; first ML hire experience; human-in-the-loop annotation/QC at scale; OCR and document understanding on low-quality real-world documents; mentoring experience.
