Data Scientist, NLP

Datavant is a data platform company and the world’s leader in health data exchange. Our vision is that every healthcare decision is powered by the right data, at the right time, in the right format.

Our platform is powered by the largest, most diverse health data network in the U.S., enabling data to be secure, accessible and usable to inform better health decisions. Datavant is trusted by the world’s leading life sciences companies, government agencies, and those who deliver and pay for care.

By joining Datavant today, you’re stepping onto a high-performing, values-driven team. Together, we’re rising to the challenge of tackling some of healthcare’s most complex problems with technology-forward solutions. Datavanters bring a diversity of professional, educational and life experiences to realize our bold vision for healthcare.

We are looking for a motivated Data Scientist to help Datavant revolutionize the healthcare industry with AI. This is a critical role where the right candidate will have the ability to work on a wide range of problems in the healthcare industry with an unparalleled amount of data.

You’ll join a team focused on deep medical document understanding, extracting meaning, intent, and structure from unstructured medical and administrative records. Our mission is to build intelligent systems that can reliably interpret complex, messy, and high-stakes healthcare documentation at scale.

This role is a unique blend of applied machine learning, NLP, and product thinking. You’ll collaborate closely with cross-functional teams to:

• Design and develop models to extract entities, detect intents, and understand document structure
• Tackle challenges like long-context reasoning, layout-aware NLP, and ambiguous inputs
• Evaluate model performance where ground truth is partial, uncertain, or evolving
• Shape the roadmap and success metrics for replacing legacy document processing systems with smarter, scalable solutions

We operate in a high-trust, high-ownership environment where experimentation and shipping value quickly are key. If you’re excited by building systems that make healthcare data more usable, accurate, and safe, please reach out.

• 3+ years of experience with data science and machine learning in an industry setting, particularly in designing and building NLP models.
• Proficiency with Python
• Experience with the latest in language models (transformers, LLMs, etc.)
• Proficiency with standard data analysis toolkits such as SQL, Numpy, Pandas, etc.
• Proficiency with deep learning frameworks like PyTorch (preferred) or Tensor Flow
• Industry experience shepherding ML/AI projects from ideation to delivery
• Demonstrated ability to influence company KPIs with AI
• Demonstrated ability to navigate ambiguity

• Experience with document layout analysis (using vision, NLP, or both).
• Experience with Spark/Py Spark
• Experience with Databricks
• Experience in the healthcare industry

• Play a key role in the success of our products by developing models for document understanding tasks.
• Perform error analysis, data cleaning, and other related tasks to improve models.
• Collaborate with your team by making recommendations for the development roadmap of a capability.
• Work with other data scientists and engineers to optimize machine learning models and insert them into end-to-end pipelines.
• Understand product use-cases and define key performance metrics for models according to business requirements.
• Set up systems for long-term improvement of models and data quality (e.g. active learning, continuous learning systems, etc.).

• Have a strong grasp of technologies upon which our platform is built.
• Be fully integrated into ongoing model development efforts with your team.

• Be independent in reading literature and doing research to develop models for new and existing products.
• Have ownership over models internally, communicating with product managers, customer success managers, and engineers to make the model and the encompassing product succeed.
• Be a subject matter expert on Datavant’s models and a source from which other teams can seek information and…