V1 — public lab. SignalLab: raw metrics, not a score yet.
Paris · Senior Data Analyst
I am building a public lab for intelligent systems — from data to agents, toward physical AI.
Since 2014 I have worked as a Data Analyst, Data Scientist and then Senior Data Analyst: banking, retail, research, public sector. The chain is always the same — understand the need, prepare the data, analyse, deliver — with machine learning when there is a case.
I am currently Senior Data Analyst at AdvanThink, on data projects for public organisations including the Ministries of the Interior and Labor. Before that, four and a half years at Ipsos, where I coordinated 2–3 people (juniors / apprentices). Since 2021 I have been mentoring data students at OpenClassrooms, in parallel with employment.
This Lab is where the next chapter is public: copilots, RAG, measurable watch, and Physical AI as a direction seen from data.
Find me
Professional
Jan 2023 — present
AdvanThink (ISoft)
Data projects for public organisations, including the Ministry of the Interior and the Ministry of Labor. ETL, analysis, ML where relevant, dashboards.
Python · SQL · Power BI · Superset · DigDash
Professional
Mar 2021 — present
OpenClassrooms
Mentoring students on data tracks, in parallel with employment.
Python · R · Tableau
Professional
May 2018 — Nov 2022
Ipsos
Preparation, statistical analysis, data viz. Coordination of 2–3 people, including juniors and apprentices.
Python · R · SAS · Tableau · Power BI
A bilingual system to observe, learn, experiment and build. SignalLab already feeds Observe with raw metrics. Decision Copilot and Knowledge Copilot are the first public agentic systems. The rest is being written here.
2021
Public data-engineering projects — Spark/S3 data lake, Redshift warehouse. Start of OpenClassrooms mentoring.
2022
Training
Alyra — Ethereum and Solidity.
2025
On-chain analytics: CSV dashboard, anomaly detection, MLflow.
2026
Agentic copilots, grounded RAG, and this Lab.
A tabular analytics agent — profiling, investigation, explainable prediction, scenarios — not a chat-over-CSV demo.
A single triage agent — closed-set classification, extraction, score, human review. No irreversible external actions.
A grounded document assistant — retrieval before generation, visible excerpts, confidence signals. Not a generic chatbot.
An on-chain dashboard from CSV exports (Dune or otherwise) — KPIs, series, cohorts. Blockchain as a data terrain, not the Lab’s identity.
UCI digits 8×8, frozen split, test accuracy: majority 10.03% vs centroid 92.48%. Not a VLM.
Entry lab: map VLM / VLA / ROS2 / simulation. The first pixel run is the 8×8 digits lab.
v2, same 20 cases: classification 100%, false autonomy 15% (down from 60% in v1).
Versioned A/B/C run: 8 rows, tool, extracted claims. B error = 0. Frozen trace: C = A.