ELGOTE — Independent AI & Data Engineer · Paris

Ismaïl-Ayoub
El Gote.

Data pipelines, analytics & applied AI.

I turn scattered data into decisions a team can trust — tested at every boundary, replayable end to end.

From the lab

Every morning, a pipeline reads France's business register.

Radar ingests BODACC open data, cleans it through a medallion architecture and answers one question: which companies just raised capital, and where. This map is its output.

companies · 30 days: 4,700
daily run · Dagster: 06:00 Paris

How it's built →

BODACC · snapshot 02.09 → 01.10.2026 · incl. 105 overseas, not drawn
Boundaries: IGN Admin Express (Licence Ouverte)

01 — The Lab

Selected work.

Three projects, code public.
The full list lives in The Lab.

All 6 projects →

02 — Articles

Latest writing.

Notes from the build.
What broke, what I measured, what I'd do again.

Oct 2026 · Data engineering · Dagster · dbt · 5 min

The partition is the date

Building a replayable daily pipeline on France's business register with Dagster, dbt and DuckDB — and how one design decision deleted a whole class of date bugs.

Read the article →

May 2026 · RAG · LLM · pgvector

A hallucinated tax threshold is worse than no answer

What building RAGuette — a citation-only RAG assistant for French freelancer admin questions — taught me about grounding, provenance, and free-tier engineering.

Jun 2025 · Security · Graph ML · PyTorch Geometric

Attacks are relational — your detection should be too

Detecting credential stuffing in real production auth logs with Graph Auto-Encoders: why modeling logins as a graph beat row-by-row analysis, and what a 0.37 → 0.93 AUC jump taught me.

A data problem that has been dragging for months? Let's talk.