Bruno Stordeur | Data Scientist

About

Turning data into decisions

I'm a Data Scientist & ML Engineer based in Paris, with a strong background in applied mathematics and computer science. Over the past few years I've delivered data and ML projects across banking, healthcare and payments — from fraud scoring and operational-risk analysis to synthetic time-series generation and clinical event prediction.

I like owning the full lifecycle: exploring and cleaning data, engineering features, building machine learning and deep learning models, and shipping them to production with solid MLOps practices — testing, CI/CD, containers and monitoring.

Outside of work, I build ambitious side projects — an automated crypto trading system and a large-scale ML research platform — to keep sharpening my data engineering and applied-ML skills.

4+

Years in data & ML

2

MSc — Data Science & Applied Maths

5

Professional certifications

3

Industries: banking, health, payments

Experience

Where I've worked

Data science & analytics roles across payments, healthcare and banking.

Cartes Bancaires

Mar 2024 – Mar 2025 · Paris

Data Scientist — Freelance

Scoring & fraud detection on EPI (Wero) transactions

The European Payment Initiative (EPI) and its Wero wallet aim to build a unified pan-European payment system. In this context, fraud prevention and detection are essential to ensure security and trust in these new payment methods.

Responsibilities

Defined a data catalog to structure P2P and e/m-commerce data, and helped integrate new catalogs into existing systems.
Created "cold start" rules to address the absence of initial data.
Monitored activity and scoring during the launch on IBM Safer Payments.
Explored and analyzed data, and developed new scoring rules using ML models and banks' fraud feedback.
Conducted technology watch and presented articles on the subject.

Python
SQL
PySpark
IBM Safer Payments
Apache Zeppelin
Hue

CHU de Brest

May 2023 – Nov 2023 · Paris

Data Scientist

Synthetic temporal data & prediction of intracranial hypertension events

Explored how machine learning and deep learning can help physicians make better treatment decisions for intracranial hypertension: generating high-quality synthetic data, predicting arterial and intracranial pressure, and predicting hypertension events.

Responsibilities

Conducted a technology watch and synthesized the state of the art.
Used a deep learning model (DGAN) to generate synthetic data, with DTW metrics to filter its quality.
Handled data extraction, cleaning, preparation and feature engineering.
Performed step-by-step prediction with deep learning and binary classification with ML models.

Python
TensorFlow
Keras
Sklearn
Pandas
Docker

BNP Paribas

Oct 2022 – Apr 2023 · Paris

Data Analyst

Historical incident analysis — Risk ORM

Within the Risk ORM division, operational risk management is crucial to monitor and mitigate risks across all banking activities. Analyzing historical incidents helps identify areas of vulnerability.

Responsibilities

Analyzed and explored data from historical incidents.
Ran correlation analysis between incidents, causes, risk types and failing business units to spot high-risk areas.
Cross-referenced incidents with implemented controls to determine at-risk perimeters.
Simplified and presented the results.

Dataiku
Python
SQL

Digitrips

Sep 2021 – Sep 2022 · Paris

Data Scientist

Cache optimization & flight-booking prediction

Flight bookings were systematically cached regardless of relevance, causing significant cost and latency. I worked on two parallel projects: classifying the relevance of flight searches to free up cache, and forecasting future flight bookings.

Responsibilities

Extracted data from SQL databases; ran analysis, exploration, feature engineering and selection.
Trained and optimized machine learning / deep learning models.
Built time-series forecasts for flight bookings (ARIMA, SARIMA, LSTM).
Simplified and presented the results.

Python
SQL
TensorFlow
Keras
ARIMA / LSTM

Projects

Things I've built

A mix of ambitious personal systems and end-to-end ML projects. The first two are original work I designed and built from scratch.

Flagship

KrakenBot — Automated BTC Trading

A production-grade automated trading system for the Kraken exchange, built as two independent 24/7 services with real-time market data ingestion, a multi-strategy engine and built-in risk management — deployed to a VPS via CI/CD.

Python · asyncio
PostgreSQL / TimescaleDB
Dash
Docker
GitHub Actions

Source

Original

Solana Memecoin ML Platform

An ML research platform analyzing 30,000+ Solana memecoins to discover behavioral archetypes and predict short-horizon price moves from the first minutes of trading — with rigorous, leakage-safe time-series methodology (walk-forward CV, per-token scaling).

Polars
PyTorch · LSTM
LightGBM
XGBoost
Streamlit

Source

Student Performance — Azure Deployment

An end-to-end ML regression app (predicting student exam scores) packaged with Docker and deployed to Azure — pushed to Azure Container Registry, served as an Azure Web App, with GitHub Actions handling continuous deployment. The focus here is the deployment pipeline.

Docker
Azure
GitHub Actions
Flask
scikit-learn

Source

Kidney Disease Classification

An end-to-end deep-learning pipeline classifying kidney CT scans, with experiment tracking (MLflow / DagsHub), data & pipeline versioning (DVC), and a Flask prediction app — a complete, reproducible MLOps template.

TensorFlow / Keras
MLflow
DVC
Flask

Source

ML Pipeline with ZenML

A production-style ML pipeline predicting customer-satisfaction scores on the Olist e-commerce dataset, with continuous training/deployment (accuracy-gated), MLflow tracking and a Streamlit app consuming the live model service.

ZenML
MLflow
scikit-learn
Streamlit

Source

SQL Spaced-Repetition Trainer

A spaced-repetition web app to practice SQL queries against a DuckDB engine, checking your answer against the expected result, with a GitHub Actions code-quality CI pipeline — a self-directed tool for keeping SQL sharp.