Take AI products from problem framing to production — scope the use case, choose the approach (generative AI, classical ML, retrieval, agents, or a hybrid), and ship it
Design prompts, tools, and retrieval strategies that work reliably in a clinical context, with structured outputs and clear failure modes
Build the services and APIs that expose AI capabilities to our products, with real attention to latency, cost, and observability
Shape the data behind each product — sources, schemas, labelling, and quality checks — and build the pipelines that prepare it for training, evaluation, and inference
Define how quality is measured (offline metrics, golden sets, human review, and live monitoring) and use the results to drive iteration