About François
French
Native or bilingual
English
Fluent
German
Conversational
Experience
- LinkupMember of the Technical StaffTECHFebruary 2026 - Today (6 months)Contributions to the 'agentic retrieval' part of the search engine:
- Implementation of a learning-to-rank stack capable of easy iteration
- Challenging the agentic methodology for deriving a query from a complex user request
- End-to-end index vectorization: benchmarking embedding models on business corpus, writing RFC, and implementing the vectorization stack
- LeboncoinMachine Learning Engineer, Search Engine TeamMay 2024 - February 2026 (1 year and 9 months)Paris, FranceData scientist specializing in information retrieval, I am part of the team that maintains and improves the platform's search engine. My achievements:
- LLM pre-processing of listings to increase overall recall: keyword completion for short listings, LLM as SPLADE for sparse vector retrieval, PoC for 'anti-SEO' document cleaning
- Reranking overhaul for the 'auto/moto' vertical
- Creation of a judgment list generation pipeline tailored to Le Bon Coin's needs
- Participation in a vector DB benchmark by setting up a Vespa cluster at Le Bon Coin's traffic scale
- Contribution to the ML guild with technological watch
- QwantData Scientist / Machine Learning EngineerSeptember 2019 - January 2024 (4 years and 4 months)Paris, FranceSteering the creation and maintenance of QwantML, an internal machine learning library for industrial use• - Implementation of the data and model lifecycle (MLFlow, DVC)• - Implementation or integration of state-of-the-art ranking models (JAX, PyTorch, LightGBM) Creation, deployment, and daily use of the main information retrieval pipeline on a Vespa index:• - Creation and cleaning of information retrieval datasets (PySpark, K8S, ArgoWorkflow)• - Fitting ranking models via QwantML and deployment on Vespa (ONNX, LightGBM)• - Creation of a new rank1 method, deployment, and paper submission at BerlinBuzzWord23 (JAX)• - Quality measurement, reproducibility, interpretability of models (LLM, MLFlow, Hydra, Captum) Contribution to an ML/NLP pipeline for assigning a query to a relevant topic index out of 1k indexes• - Creation and cleaning of query datasets with their topics (PySpark, ArgoWorkflow)• - Vectorization of queries and vector matching against 1k indexes (HF, Vespa)• - Construction of meta-indexes representing the acceptable query scope for each index (FAISS) NLP exploration to increase ranking recall on a 10B docs index via document expansion• - Mentoring an M2 internship on the industrialization of sparse+dense NLP ranking (HF, SPLADE)• - Document categorization and document expansion (HF, SPLADE) Explorations on RAG with hybrid sparse/dense search (Llama-index, Vespa, RAG, LLM) Steering a project to maintain a satisfactory quality level of the main Qwant index• - Training and deployment of a spam removal score for 5B items in the index (SKLearn, HF, Vespa)• - Mentoring and deployment of an M2 internship on prioritizing documents for crawling (SKLearn)• - Development and deployment in containerized environments using Docker, Kubernetes, and ArgoWorkflow
Recommendations
Be the first to recommend François
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- PhD, Mathematics and StatisticsTelecom ParisTech2017Doctorat, Mathématiques et statistiques
- Engineering Diploma from ENSAI, Statistical EngineeringEcole nationale de la Statistique et de l'Analyse de l'Information2013Diplôme d'ingénieur de l'ENSAI, Génie Statistique