System design & architecture

Python/FastAPI microservices with lab and project isolation, OIDC authentication, fine-grained access control and per-project compute environments. Designed as a stable base, so downstream services integrate through one consistent model instead of special cases.

Migration of scientific data from the legacy system into the new core service: schema mapping between the old and new domain models, scripted and re-runnable backfills, validation passes that compare counts and key fields, and a staged cut-over by data type while the legacy path stayed available.

Service boundaries, domain models and versioned API contracts for the core entity service. Each model has one owner; other services read and write through the contract, never through the database.

Separation between the interactive UI, an API layer for launch, status and results, and parallel compute behind it. Results are streamed back to shared sessions, so several researchers can follow one run in real time.

An app shell for routing, authentication and layout, with product modules built and deployed independently on top of it. Module boundaries follow product areas, so each team ships without coordinating a full release.

Shared backend building blocks for authentication, validation, notifications and audit logging, reused by every module instead of being rebuilt per feature.

A permission model for staff roles and administrators, enforced on the API and reflected in the UI, so each user sees only the data and actions their role allows.