Creating APIs That Data Scientists Will Love
summary
Five concrete requirements for data-science-friendly APIs: publish a Python SDK on PyPI so code like pip install and client.list_players returns pandas-ready frames; include industry external IDs (example: GSIS ID) so joins with external sources (NFL Data API) work; define and validate precise types using FastAPI + Pydantic to generate OpenAPI contracts; offer a bulk-download endpoint for full dataset exports; support querying by a technical last_changed timestamp for efficient incremental Airflow pipelines.
AI-written summary · about
chapters (9)
- 0:00API design for data scientists
- 1:30Data scientist workflows
- 5:05Five tips for APIs
- 6:12Python SDK
- 11:03Standard external IDs
- 15:00Strict data types
- 17:28Bulk downloads
- 20:25Query by last change
- 22:10Summary and call to action
related talks
46mCreating Solid APIs
28mData Science Design Patterns
47mBuilding APIs your users will want to use
33mBuilding Data Science Tools for Sustainable Transformation
24mAPIs and Data products: How do they impact your business model?
26mServing Data as API
26mData with a mission: a COVID-19 API case study
42mDevelopment of Data Science Ecosystem for Ruby