Full-time B2B | Remote Poland | US Healthcare Real-World Data Context
We are an AI-native data and technology partner for private capital and healthcare. Founded in 2010 and headquartered in Warsaw, we work with leading PE firms, VC funds, and healthcare organizations to build proprietary data infrastructure, deploy AI solutions, and drive AI-native transformation.
Our clients manage a cumulative $1.2T+ in assets. Our average engagement runs five years. Our NPS sits above 80. We don't need to claim credibility – we can show it.
We've also done to ourselves what we now do for clients. We've restructured our own company around AI – tools, policies, roles, delivery models. This isn't a pitch. It's a playbook we've already run, and we're hiring the engineers who will run it for others.
The Opportunity
Our client is a US healthcare data company. They turn real-world EHR data from specialty registries (ophthalmology, urology, oncology) into research-grade data products used for real-world evidence.
They are bringing a clinical-data extraction product to market and need extra senior capacity in their Data & Science team. The core problem: getting reliable, structured facts out of messy clinical notes with language models, and proving they are reliable. Your outputs feed research studies and commercial data products, so accuracy and explainability matter as much as shipping.
You'll join the client's Data & Science team (the role is called Senior Quantitative Scientist on their side) and work every day with their Product and Medical teams on the US East Coast.
What you'll be doing
The core of the work is applying language models to unstructured clinical text: building extraction pipelines, measuring how well they work, and taking them to production. You'll work with Medical and Product to turn clinical questions into concrete model requirements, and explain results to people who don't read confusion matrices. We're looking for someone who combines research depth with production instinct.
Your Responsibilities
- Build and improve LLM- and transformer-based pipelines that extract structured data from clinical notes: named entity recognition, text classification, relation extraction
- Evaluate models rigorously: gold-standard datasets, error analysis, metrics that clinicians and researchers trust
- Fine-tune, deploy and monitor models in production
- Work with Medical and Product teams to turn clinical questions into concrete data and model requirements
- Present results and trade-offs clearly to technical and non-technical audiences
Requirements
- Master's degree or PhD in a quantitative field (computer science, machine learning, data science, biostatistics, or similar). A PhD is welcome but not required
- Around 5 years of hands-on data science or ML work, including models you took to production on messy real-world data
- Hands-on NLP with transformer encoders: fine-tuning BERT / RoBERTa-family models (including biomedical or clinical variants such as BioBERT or ClinicalBERT) for NER, classification or relation extraction, typically with Hugging Face Transformers
- Hands-on work with generative LLMs: prompt design and structured outputs, fine-tuning (for example LoRA / PEFT), RAG, and evaluating LLM outputs at scale (gold sets, error analysis, LLM-as-judge)
- A rigorous approach to evaluation – you know how to tell whether a model is actually right, not just plausible
- Strong Python and SQL
- AI-native way of working: you use tools like Claude and Cursor as a core part of how you deliver
- Fluent English and confidence presenting to cross-functional and non-technical stakeholders
Strongly preferred:
- Experience with EHR data or clinical notes, and familiarity with coding systems such as ICD, CPT or RxNorm
Nice to Have
- PySpark and Databricks (including MLflow)
- Training, serving and monitoring models on AWS, especially Amazon SageMaker
- R or TypeScript – R is part of the team's analytics stack, and some legacy code is written in TypeScript
- Ophthalmology, urology or oncology data
- Medical imaging data
- Real-world evidence studies or peer-reviewed publications
Benefits
- Unrestricted AI Stack & Premium Gear: Fully paid licenses for Cursor, Claude Pro, etc.
- Total Autonomy (Remote-First): No filler meetings, no Jira bloat, no micromanagement. You own the workflow. We care about shipped systems in production, not logged hours.
- Direct Impact: You'll work face-to-face with our CEO, CTO & VPs and directly with client leadership.
- Frontier Engineering Culture: Build alongside elite engineers who are shipping systems that drive real business decisions. Backed by continuous growth and a strong knowledge-sharing culture (check our YouTube).
Sounds like a perfect place for you? Don't hesitate to click apply and submit your application today!