Project · ingestion
A reproducible ingestion layer for messy epidemiological reality.
Turns inconsistent, multi-source surveillance feeds into analysis-ready data — so every downstream conclusion is traceable to clean inputs.
The problem
[What was messy — sources, formats, the failure modes you had to absorb.]
What I built
[Schema enforcement, validation rules, logging, idempotent runs. Name the stack.]
Artifacts
[Repo link · README · sample run · before/after of a dirty record.]
Why it matters here
This is the root of the analytical DAG — transformations, inference, and forecasting all inherit their trustworthiness from this node.