Your Name

About

I build software that keeps working: backends, the pipelines that ship them, and the AI features on top.

• Location: City, Country
• Languages: English, French
• Availability: Open to new roles from January 2027

Skills

Experience

1. Senior Engineer, Platform & AI at Northwind Labs

   2023-01 — present · Remote

   Own the deploy pipeline and the AI features that ride on top of it.

   • Cut release time from a day of manual steps to 20 minutes, self-serve
   • Built the canary rollout that now gates every production deploy
   • Shipped an internal search assistant used by support and sales

   Technical details

   Stack: Kubernetes, Terraform, GitHub Actions, Python, AI/retrieval

   Releases used to mean a shared spreadsheet, a Slack thread, and someone awake at 6am. I replaced the manual steps with a GitHub Actions pipeline that runs the test suite, deploys a canary, watches error rates for 15 minutes, and only then promotes to the rest of the fleet — with an automatic rollback if the canary’s error budget is exceeded. On the AI side, I built a retrieval-augmented search assistant over the internal docs that support now uses as a first line of triage.

2. Software Engineer, Data Platform at DataLoom

   2020-07 — 2022-12 · Berlin, Germany

   Built the tools the data team used to test and ship pipelines without breaking production.

   • Replaced ad-hoc notebooks with a tested, scheduled pipeline framework
   • Cut a report that took a day to generate down to 20 minutes
   • Onboarded 6 analysts onto the new framework with no incidents

   Technical details

   Stack: Python, Airflow, dbt, Snowflake

   The data team shipped pipelines straight from notebooks, so a bad join could silently corrupt a report for weeks before anyone noticed. I introduced a small internal framework with schema checks, a local test runner, and a staging environment that mirrored production data volumes. Once analysts could run their own pipeline in staging before merging, review time and incident count both dropped.

3. Backend Engineer at Globex Bank

   2018-03 — 2020-06 · Remote

   Rebuilt a payments backend so it stopped losing transactions during peak load.

   • Cut duplicate-payment incidents from weekly to zero over 8 months
   • Moved the settlement job from a nightly batch to an event stream
   • Wrote the on-call runbook the team still uses today

   Technical details

   Stack: Java, Spring Boot, PostgreSQL, Kafka

   The settlement pipeline ran once a night and any failure meant re-running the whole batch by hand. I split it into idempotent, replayable steps driven off a Kafka topic, added a dead-letter queue for the rows that needed a human, and instrumented every stage so on-call could see exactly where a run had stopped. Transaction volume tripled over the following year without another rewrite.

Education & certifications

Blog

• Grounding an assistant in docs that change weekly

  2025-06-18 · 1 min read

  The hard part of a docs assistant isn't the model. It's keeping the index honest.

• Canary deploys that roll back themselves

  2025-03-10 · 1 min read

  How a small error-budget check turned our worst deploy days into a non-event.

• What a data pipeline test actually needs

  2024-11-02 · 1 min read

  Notebooks don't fail loudly. Here's the smallest test setup that makes them fail loudly.

Contact

I usually reply within 2 working days.

• github
• linkedin