About Me
Data Scientist with a double engineering degree from ENSAI Rennes (Data Science) and ENSAE Dakar (Statistics & Economics), I combine a rigorous background in statistics and econometrics with practical experience in ML mo…
Data Scientist with a double engineering degree from ENSAI Rennes (Data Science) and ENSAE Dakar (Statistics & Economics), I combine a rigorous background in statistics and econometrics with practical experience in ML modelling and production deployment. I designed and industrialised full ML pipelines at Little John (insurtech, Rennes) and conducted applied research at CREST-ENSAI. Available immediately.
Experience
ML Engineer & Data Scientist
<p>Fine-tuned Skywork-Reward-V2-Qwen3-0.6B (0.6B params) on an insurance document corpus (certificates, claims, passports, bank details...)</p><p> Achieved 86.4% accuracy vs 74.1% for GPT-4o-mini</p><p> Built a model 15× smaller with local sovereign deployment</p><p> Improved inference speed to 1.76 s/doc vs 3.78 s/doc</p><p> Achieved F1-score of 85.3% vs 70.3%</p><p> Built a full training pipeline (JSONL), stratified split, and published it on HuggingFace Hub</p><p> Designed a multi-agent AI architecture using Pydantic AI</p><p> Created YAML business templates per broker module</p><p> Used AG-UI for real-time agent-broker interaction</p><p> Parallelised execution via Ray</p><p> Exposed services with Ray Serve</p><p> Reduced processing time from 1 week to under 1 hour</p><p> Built an ingestion pipeline via Azure Document Intelligence (OCR) to structured JSON and Azure Blob Storage</p><p> Developed a RAG system with ChromaDB vector store and OpenAI embeddings</p><p> Implemented a multi-model fallback</p><p> Industrialised the solution with Docker, CI/CD, FastAPI, Logfire monitoring, and GDPR compliance</p>
Research Intern — Machine Learning
<p>Applied Topic Modeling (LDA) for automatic classification of technical skills</p><p> Performed text preprocessing</p><p> Built reproducible pipelines</p><p> Conducted econometric modelling of the skill–wage relationship</p><p> Used cross-validated feature selection</p><p> Produced methodological reports</p><p> Presented results to an academic audience</p><p> Received the highest grade of the cohort for methodological rigour and originality</p>
Statistical Studies Officer & Platform Developer
<p>Developed and deployed an HR/recruitment management platform</p><p> Improved process efficiency by 30%</p><p> Exploited and structured large administrative databases</p><p> Automated data processing workflows</p><p> Produced institutional statistical indicators and reports</p>
Statistics Consultant
<p>Performed econometric analyses for impact evaluation studies using panel data, regression, and causal inference</p><p> Cleaned and structured survey data</p><p> Produced statistical reporting</p><p> Drafted technical notes and reports for institutional and international clients</p>
PROJECTS
Consulting Company activity analysis
<p>Analysis the activity of the company in terms of recruitment, to know entries and outs of employees using classic time series (ARIMA, SARIMA) and Facebook prophet package (2017).</p>