Staff Data Scientist | NLP

Machinify is a leading healthcare intelligence company with expertise across the payment continuum, delivering unmatched value, transparency, and efficiency to health plan clients across the country. Deployed by over 85 health plans, including many of the top 20, and representing more than 270 million lives, Machinify brings together a fully configurable and content-rich, AI-powered platform along with best-in-class expertise. We’re constantly reimagining what’s possible in our industry, creating disruptively simple, powerfully clear ways to maximize financial outcomes and drive down healthcare costs.

We're hiring a Data Scientist focused on natural language processing to build models that turn unstructured text into product features and business insight. You'll own problems end-to-end — framing, data, modeling, evaluation, and shipping — and work closely with engineering and product to put your work in front of users.

What you'll do

• Design and train NLP models for tasks like classification, entity extraction, retrieval, summarization, and semantic search
• Fine-tune and evaluate LLMs (open-source and API-based); build RAG pipelines and agentic workflows where appropriate
• Build robust evaluation harnesses — offline metrics, human-in-the-loop review, and online A/B tests
• Partner with ML engineers to productionize models (latency, cost, monitoring, drift detection)
• Turn ambiguous product questions into well-scoped ML problems and communicate tradeoffs clearly to non-technical stakeholders

What we're looking for

• 3+ years of applied ML experience with a meaningful portion in NLP
• Strong Python and the modern NLP stack: PyTorch or JAX, Hugging Face Transformers, spaCy, sentence-transformers
• Hands-on experience fine-tuning transformer models (LoRA/QLoRA, instruction tuning, preference optimization) and/or building production RAG systems
• Solid grounding in evaluation: knows the difference between BLEU/ROUGE/BERTScore/LLM-as-judge and when each is misleading
• Comfortable with SQL, vector databases (pgvector, Pinecone, Weaviate, or similar), and one major cloud (AWS/GCP/Azure)
• Clear written and verbal communication; can defend a modeling choice and also admit when a heuristic beats a model