Role summary: The overall technical lead and architect. Designs the metadata schema, builds the simulation onboarding pipeline, deploys metadata embedding pipeline and OpenSearch k-NN vector store, and authors data export format spec for AI/ML use case. This role is the deepest technical seat on the engagement:
Key responsibilities Run the Sprint 1 architecture review of the existing UAT codebase (S3 + Glue + S3 Tables + OpenSearch + Athena) and deliver written gap findings. Design the metadata schema, taxonomy, and field catalogue (Light, Brain, Power). Tune data orchestration — Glue jobs, Athena queries, S3 Tables config, scheduling. Lead the deep-dive technical sessions with analysts on visualization requirements Build and validate the simulation data onboarding pipeline against real data — including the 30 GB-per-run acoustic spectra dataset. Configure and validate the OpenSearch k-NN vector store and the Bedrock embedding pipeline. Author the AI/ML data export format specification and the AI onboarding pattern document. Co-design the API middleware blueprint with the Cloud Infrastructure Architect.
Must-have Principal-level hands-on data engineering on AWS — 7+ years Deep production experience with S3, S3 Tables, Glue, Athena, and OpenSearch (including k-NN / vector search) Built and shipped vector embedding workloads Strong metadata modelling and data taxonomy design experience for scientific or engineering domains Comfort working with Parquet, JSON-LD, and large binary scientific data formats (mesh, time-series, spectra) Python proficiency; PySpark / Glue job tuning experience
Nice-to-have / differentiators Prior simulation / CAE / HPC data lake experience (Ansys, Siemens NX, BETA CAE, OpenFOAM, etc.) Familiarity with surrogate model training data pipelines Experience with SageMaker Unified Studio or comparable governed data-mesh tooling (in case of required integration) Multi-cloud data engineering (AWS GCP) experience Published or contributed to AWS data architecture patterns or blueprints…