Paris, Île-de-France, France
Machine Learning Engineer at Lifen, working on deploying production-grade LLM systems for healthcare. I specialize in building end-to-end AI solutions, from model design and fine-tuning to deployment and monitoring at scale. My work focuses heavily on Natural Language Processing, particularly LLMs applied to complex, unstructured clinical data, as well as MLOps and AI infrastructure. At Lifen, I’ve prototyped and deployed fine-tuned LLMs to automate the structuring of clinical documents, designed internal MLOps tools leveraging MLflow and Kubeflow, and contributed to research on unstructured clinical data processing. Previously, I worked at C3 AI as a Solutions Engineer, where I deployed RAG applications over large document corpora, collaborated closely with data scientists on enterprise ML systems, and delivered AI solutions with measurable business impact. I’m particularly interested in applied AI, LLMs in production, and AI systems with positive social or healthcare impact.
Realizations: - Prototyped then deployed fine-tuned LLMs to automate the structuration of complex clinical data - Designed a custom suite of internal MLOps tools, leveraging open-source components (MLFlow, Kubeflow, Dash) and spearheaded the adoption process - Contributed to the redaction and publication of a breakthrough article on unstructured clinical data processing
- Deployed and optimized a RAG application containing over 10,000 financial documents - Assisted our Data Science team in automating an IPO filing workflow for a major law-firm, using LLMs
- Improved and maintained a predictive maintenance application saving over $500k annually for a pharmaceutical group - Transformed customers' needs and requirements into concrete software solutions, for pharmaceutical and state applications - Collaborated with a team of Data Scientists to optimize and deploy ML models for Enterprise AI solutions
For 5 months as a Research Intern within the Corail Research Group, I worked on SeaPearl: an open-source hybrid Constraint Programming (CP) solver improved with RL and coded in Julia. Realizations: - Prototyped a new architecture of SeaPearl optimized for GPUs and gained a 10 times speedup; - Conducted research to train graph-based Deep Q Networks (DQN) to efficiently solve CP problems, reducing search tree size by a factor of 3;
For 3 months as a Software Engineer Intern at LINAGORA, I developed a SaaS service enabling transparent access to a NoSQL database, paired with Twake co-founder. Realizations: - Developed a back-end service to abstract an encrypted and replicated database (inspired by Google FireBase) for a chat application, using NodeJS, TypeScript, and Apache Cassandra, with a live performance summary; - Familiarized with SCRUM development over 12 weeks and 6 sprints;
During the Ecole Polytechnique's initial training, I served 5 months as a Cadet in an infantry regiment in French Polynesia. Missions: - Ordered at a group of 6 during a 4-week surveillance mission; - Seconded a Lieutenant in leading a 3-week humanitarian mission. Supervised 3 worksites autonomously; - Pitched water management solutions based on data analysis of a remote military position water supply, and improved water and jet-fuel logistics by implementing automated logging;