AI Engineer

POSITION SUMMARY

Natera is seeking a Principal AI Engineer to set technical direction for, and hands-on build of, the Vector Store GenAI Platform that powers document intelligence and retrieval across our Real-World Data (RWD) organization. The platform ingests clinical documents from multiple sources, indexes them for high-quality search, and makes that content available to downstream AI workflows and business use cases across oncology and other clinical domains.

This is a principal-level individual contributor role with architecture-plus-build ownership of retrieval, document processing, and ingestion systems. You will define platform strategy and sequencing, ship production code in a cloud environment, raise the reliability and quality of search and retrieval, and prove new approaches through POCs and benchmarks. You will partner with clinical, engineering, and product leaders to turn ambiguous priorities into a clear technical roadmap, establish standards for how retrieval serves agents and re-extraction workflows, and deliver measurable improvements in platform quality and operability—with high autonomy and limited day-to-day direction.

The ideal candidate combines deep hands-on engineering with a track record of owning complex AI platforms end to end. You set direction, implement critical path yourself, and elevate the technical bar across the teams that consume and extend the platform.

RESPONSIBILITIES

• Own platform architecture and hands-on delivery for the Vector Store GenAI Platform, including search and retrieval, document processing and chunking, ingestion pipelines, and APIs that serve downstream AI workflows.
• Set technical direction for hybrid search and retrieval quality; evaluate embedding models and retrieval approaches through working implementations, POCs, and benchmarks, and drive adoption of what works in production.
• Advance document parsing and chunking so clinical content is indexed in a way that supports accurate, structured, and faceted retrieval across common document types.
• Define and build stable, reproducible retrieval for agent and re-extraction workloads, including durable content identifiers, flexible context assembly, versioned indices, and quality or confidence signals.
• Strengthen document classification and ingestion visibility so source material is discoverable, operable, and measurable at scale across historical and ongoing data flows.
• Continuously improve retrieval quality through better deduplication, clinical concept normalization, and exploration of promising retrieval methods where they add clear value.
• Integrate LLM capabilities into existing production systems with attention to performance, scalability, cost, observability, and safe operation in a regulated environment.
• Partner with clinical AI, engineering, product, and SME stakeholders to prioritize workstreams, define success metrics, and sequence roadmap investments without disrupting the live platform.