Applied Healthcare Researcher

Protege

About The Position

We are hiring Applied Healthcare Researchers to join a team within DataLab focused entirely on healthcare training data. Our customers are researchers at the frontier labs and AI startups building specialized healthcare models. They come to us with model-development problems, not dataset specifications. Figuring out which healthcare data actually solves their problem, and proving that it does, is the research question we answer in DataLab. In this role you will work directly with researchers at those labs to understand what they're trying to train or evaluate, determine what healthcare data can support it, and do the research needed to demonstrate that it will. This is fast-iterating, customer-facing research on a customer's timeline. You will be the primary technical and research link to the customer — not a technical resource brought in for credibility, but the person driving the conversation and pulling in the solutions, engineering, and data partnerships teams as needed.

Requirements

Advanced degree (PhD or Master's plus 3+ years industry experience) in machine learning, computer science, biomedical informatics, epidemiology, statistics, or a related quantitative field — or equivalent applied experience.
Hands-on experience building and evaluating ML or LLM-based systems for extraction, classification, or prediction on real-world data.
Experience working with healthcare data: claims, EMR/EHR, clinical notes, imaging, registries, or similar. You understand why real-world clinical data is messy and what that means for model training.
Strong Python and SQL, with the ability to work independently against large datasets.
Experience designing evaluations — measuring data quality and dataset representativeness.
Demonstrated ability to work directly with technical stakeholders and translate ambiguous goals into concrete, defensible research plans.
Comfort operating on a customer's timeline without lowering the standard of the research.

Nice To Haves

Is energized by working directly with customers, and specifically by working with other researchers as peers.
Moves fast on messy, real-world problems and knows which corners can and cannot be cut.
Is rigorous about what the data can and cannot support, and willing to tell a customer when the answer is no.
Enjoys the full arc — scoping a vague problem, doing the research, and showing the result to the person who asked for it.
Thinks about leverage: builds the reusable version rather than the one-off when it's worth doing.

Responsibilities

Serve as the primary technical and research point of contact for healthcare customer conversations.
Translate a lab's model-development goals into concrete, feasible data strategies.
Help customers scope opportunities and identify the highest-value data available to them.
Explain data limitations, tradeoffs, and potential biases to technically sophisticated stakeholders while grounding conversations in what real-world data actually looks like.
After delivery, answer the research questions customers raise about the data we provided. Delivery is not the end of the relationship.
Develop and evaluate methods — fine-tuning, LLM-based extraction, classification, rules-based approaches, or whatever the problem calls for — to demonstrate that a dataset can support a customer's training or evaluation objective.
Design and run feasibility research pre-contract: can this data support this model objective, at what quality, with what caveats.
Build the evidence base that makes a data strategy credible — benchmarks, validation analyses, error characterization, and honest assessments of where the data falls short.
Partner with the Assessments team on healthcare benchmarks across modalities.
Evaluate whether requested variables, labels, or cohort definitions are achievable with available healthcare data.
Identify proxy variables or alternative dataset structures when the ideal variable doesn't exist.
Analyze partner and source datasets — schema, field availability, quality, completeness, and required transformations.
Contribute to our point of view on which healthcare data matters most for which modality and which stage of model development.
Help evaluate new data partners and identify datasets worth acquiring before a customer asks for them.
Produce reusable research, evidence, and technical collateral rather than starting from scratch for each opportunity.
Identify where a successful one-off approach should become a repeatable workflow, and work with Product and Engineering to operationalize it.
Help expand proven healthcare datasets across multiple customers instead of selling them once.
Work with Solutions and FDEs from the beginning of an opportunity.
Coordinate with Healthcare Data Partnerships on sourcing and with Product and Engineering on tooling.