alysanne.dev

backend systems / production evidence

Correct systems. Unreliable data.

Backend engineering for concurrent pipelines, unreliable inputs, and the operational proof required to trust them.

Read my résuméContact

Technical diagram for outside-in production debugging
Start with the request and persisted state. Move inward only when the evidence requires it.

Operational method

I prefer observable boundaries and concrete checks over plausible explanations.

  1. 01Reproduce

    Begin at the system boundary with the smallest failing request.

  2. 02Observe

    Compare logs, SQL, Redis, telemetry, and raw responses.

  3. 03Constrain

    Revise the hypothesis until one failure mode remains.

  4. 04Prove

    Repair the behavior, then keep it fixed with a regression test.

Recent notes

View archive
  1. GPS can lie: designing telemetry validation before side effects

    A practical ordering for duplicate, time, coordinate, rate, and movement checks in an ingestion pipeline.

  2. Publishing a multi-tenant ETS index without partial state

    How composite keys, one mutation owner, and generation swaps turn a fast cache into a dependable read model.

  3. Debugging distributed data from the outside in

    Why I begin with the raw request and persisted state, then move inward toward logs and code.

Working set

Production
Telemetry ingestion, multi-tenant data paths, third-party APIs
Laboratory
Embedded Rust, local-first software, agent orchestration
Tools
Elixir/OTP, PostgreSQL/PostGIS, Redis/ETS, AWS, Docker, Linux