Lucas Potin | Data Scientist & PhD in Data Mining

About

From graph mining to applied AI

I like solving problems from data. This curiosity led me from applied mathematics to data science, then to a PhD focused on detecting corruption in public procurement through the analysis of complex graphs.

Today, I work both on predictive models and on tools built around generative AI. Technologies evolve, but what motivates me remains the same: starting from a concrete problem, exploring different approaches and building useful solutions.

Outside work, I play competitive tennis and regularly play bridge.

Journey

From applied mathematics to data tools

2016–2020

Engineering degree in mathematical engineering

Applied mathematics, modeling and optimization.

2020

MSc in Data Science

Machine learning, statistics and applied data projects.

2021–2025

PhD in Data Mining

Graph analysis and anomaly detection in public procurement.

Today

Data Scientist @ GSE

Forecasting, generative AI and decision-support tools.

Skills

Expertise & technologies

Data Science & ML

Machine Learning Statistics Graph Mining

Generative AI

LLM RAG AI Agents

Engineering & Cloud

Python SQL Azure Elasticsearch Docker

Projects

Selected work

Apps, dashboards and APIs developed around concrete data problems.

Graph Mining | Dashboard

DramaPang

Objective: Classify plays from the relationships between characters.
Approach: Build character networks and extract discriminant patterns.
Result: Interactive web application.

Graph Mining Classification NetworkX

App Code

Machine Learning | Dashboard

Retail sales prediction

Objective: Forecast weekly sales by product family.
Approach: Compare forecasting models, including Prophet and XGBoost.
Result: Interactive dashboard for visualization and comparison.

Forecasting Machine Learning XGBoost

App Code

Data engineering | API

Siretizator

Objective: Automatically match organizations to their SIRET identifier.
Approach: Data normalization and fuzzy matching against the SIRENE database.
Result: Entity reconciliation API.

Open Data Entity Resolution FastAPI

API Code

Research

Scientific publications

Work around graph mining, open data and public procurement analysis.

2025 Data in Brief

Processing and consolidation of open data on public procurement in France (2015–2023)

Consolidated database of French public procurement contracts designed to support economic analysis, research and anomaly detection.

2015–2023 period covered

~300K contracts consolidated

Authors: A. Deschamps, L. Potin

Article Code

2025 ACM TKDD

Pattern-Based Graph Classification: Comparison of Quality Measures and Importance of Preprocessing

Systematic comparison of quality measures used to select discriminant patterns for graph classification tasks.

38 measures compared

92% fewer patterns on one benchmark

Authors: L. Potin, R. Figueiredo, V. Labatut, C. Largeron

Article Code

2023 Nature, Scientific Data

FOPPA: an open database of French public procurement award notices from 2010–2020

Open, cleaned and documented database enabling large-scale study of French public procurement from TED.

2010–2020 period covered

~1.4M lots integrated

97.5% buyers identified

Authors: L. Potin, V. Labatut, P.H. Morand, C. Largeron

Article PDF Code

2023 ECML PKDD

Pattern mining for anomaly detection in graphs: Application to fraud in public procurement

Explainable anomaly detection method for graphs applied to the identification of risk-prone situations in public procurement.

15,793 patterns extracted

0.95 F1-score on FOPPA

Explainable discriminant patterns identified