Aubin Mazet

Opale Grid turns unstructured files into a live, spreadsheet-like workspace so analysts scan, sort and interrogate thousands of pages with formula-level precision. Role: Own product vision and development for Opale, driving AI-native product design and key technical decisions as the product scaled to production usage (10M+ pages processed). Key achievements: - Established an AI-native product development loop with designers and software engineers, tightly coupling product workflows, UX, and AI behavior, enabling multiple feature releases per week - Designed and shipped a state-of-the-art agentic setup embedded directly into the product - Owned end-to-end software decisions across the product, aligning core data flows, APIs, and user-facing features to support rapid iteration across the entire product surface

Quantmetry is a leading pure-play AI consulting firm in France, where I worked as a GenAI/Data Lead across multiple enterprise engagements. Key Projects: • Banking (GenAI Architect): Designed a secure hybrid cloud and on-premise architecture for GenAI deployment, validated by IT and CISO teams in a regulated banking environment. • Insurance (GenAI Lead): Defined the target architecture for conversational agents and built a robust RAG evaluation framework, later used as a reference to industrialize chatbots at scale. • Private Equity (GenAI Lead): Led a 10-engineer team to build and productionize a RAG-based internal assistant (100+ users, 1M+ documents, 1+ year in production on Azure) to boost analyst productivity. • Retail (Datalab Tech Lead): Led a 10+ person AI team and redesigned the group’s MLOps platform (CI/CD, MLflow), enabling the deployment of 8 AI projects in a Databricks and AWS environment. • Industry (Senior Data Engineer): Deployed an on-prem data lake (Kubernetes, Spark, Delta Lake, Airflow) and delivered the first production use cases, centralizing heterogeneous data and accelerating time-to-production through standardized pipelines and governance. • Healthcare (Senior Data Engineer): Built and optimized a real-time data platform (Citus, Azure, streaming pipelines) handling 5B+ rows for large-scale analytical workloads.

Context: Worked on internal financial ML projects spanning fraud detection and revenue forecasting. - Developed GAN-based auto-encoder models for anomaly detection on highly imbalanced transactional data, focusing on rare-event detection with limited labels. - Built knowledge-graph–based modeling approaches to represent financial entities and constraints, improving the structure and interpretability of forecasting models. Core stack: Python, PyTorch, GANs & Auto-encoders, Knowledge Graphs

- Lead weekly management meetings, presenting KPIs - Develop data-driven initiatives to drive growth - Coordinate cross-functional teams and projects