AI Data Engineer

Howard Hughes Medical Institute•Headquarters, KY

2d•$128,817 - $161,021•Hybrid

About The Position

The AI Accelerator exists to turn AI into daily reality across HHMI’s administrative and operational functions. This role builds the data-engineering foundation every AI application and knowledge layer at HHMI runs on. The work is hands-on. This person implements the pipelines, transformation patterns, and orchestration framework that turn HHMI’s institutional content into governed, AI-ready data. They own the day-to-day execution of the data-engineering side of the AI fabric — pipeline development, medallion curation, workflow-orchestration framework, and governance implementation — under the design authority of the Principal Knowledge & Data Architect. This role works closely with the Principal Knowledge & Data Architect (who owns the knowledge and retrieval layer’s design), the AI Developer (who consumes what this role builds), and Operations Capabilities’ Data Integration Engineer (who lands source-system data at the interface). The AI Data Engineer is the seat that connects those layers into working data flow. HHMI’s Principal K&D Architect designs the knowledge and data foundation for institutional AI, but the foundation only comes to life when someone builds and operates the pipelines that carry data through it. Without this role, our AI systems either wait on data that isn’t ready or consume content whose quality can’t be guaranteed. This role is what makes the K&D Architect’s design real day-to-day — and it’s the same role that keeps it working over years, as content evolves and models change.

Requirements

Hands-on production data engineering: at least four years designing, building, and operating production data pipelines. Not a Databricks-course track record — real production experience where you owned the pipeline through breakage, iteration, and recovery.
Databricks and Spark depth: Delta Lake, medallion architecture, Delta Live Tables, Workflows, Databricks SQL, Unity Catalog. Comfortable at the layer where code meets platform.
Python and SQL fluency: PySpark, ETL patterns, and SQL that runs at scale. Version control (Git), CI/CD for data pipelines, and infrastructure-as-code (Terraform) as working tools.
AI-adjacent data engineering: real experience building the data foundation for AI use cases — embedding pipelines, vector stores, chunking strategies, retrieval evaluation. Not required to be an ML researcher; required to have built the plumbing.
Workflow orchestration: Databricks Workflows, or Airflow in production. Retry semantics, dependency management, failure handling — as working discipline, not concepts.
Data quality and observability: Great Expectations, Databricks data-quality monitors, or equivalent. Treats data quality as a first-class engineering concern.
Governance discipline: works with Unity Catalog structures, understands sensitivity classification, and designs pipelines with access control and audit in mind from the first commit.
AWS foundations: IAM, S3, KMS at the level needed to work in a Databricks-on-AWS environment. Not required to be a cloud architect; required to be productive.
Communication: works productively with the K&D Architect on design, AI Developer on integration, and Operations Capabilities on contracts. Explains data-engineering trade-offs to non-engineers.
Education and experience: bachelor’s degree or equivalent, plus at least four years of hands-on data-engineering experience with meaningful exposure to AI or knowledge-management use cases.

Nice To Haves

Prior experience with knowledge graphs (Neo4j or comparable), entity resolution, or semantic data models.
Experience with the modern data stack alongside Databricks-native tooling.
Familiarity with LLM-based extraction, chunking, and evaluation frameworks.
Background in research, academic, or mission-driven institutional environments.
Experience with cross-platform data engineering (Snowflake, BigQuery) — cross-platform judgment is useful even when Databricks is the primary tool.

Responsibilities

Build the AI-facing data pipelines: ingestion from the landing zone Operations Capabilities delivers, transformation through raw → bronze → silver → gold, and serving of governed, AI-ready content for downstream consumption.
Implement the medallion architecture: design patterns from the K&D Architect become working pipelines, tables, and materialization schedules. Delta Lake tables designed with partitioning, optimization, and evolution in mind.
Own the workflow-orchestration framework: Databricks Workflows, Delta Live Tables, retry policies, alerting routes, run history, cost tags.
Implement governance patterns: Unity Catalog structure, sensitivity classification, access control, and audit for AI-facing data assets. Design comes from the K&D Architect; day-to-day implementation lives here.
Build and operate retrieval-supporting infrastructure: embedding pipelines, vector store maintenance, reindexing when models upgrade, retrieval evaluation frameworks.
Partner with Operations Capabilities on source-system contracts: define what the AI Fabric consumes at the landing zone — schema, cadence, SLA, quality thresholds. Own the platform-side of that contract.
Design and operate data quality: data-quality checks, freshness monitoring, drift detection, and the alerting that surfaces issues before they hit AI users.
Support AI Developer velocity: when AI Developers deploy into product teams, this role is the data engineer they turn to when a use case needs specific data.
Contribute to and consume the reference-pattern library: reusable pipeline patterns, code templates, and standards live in the shared platform layer. This role uses them, contributes new ones, and evolves them as we learn.