Marc Mayol

At CANCOM I worked as an AI Engineer on the CANCOM Assistant, a multi LLM orchestration system built with FastAPI, LangGraph and hexagonal architecture. It was an international project serving clients across the DACH region (German-speaking market), so the solution had to meet specific linguistic and operational requirements of German-speaking users. My responsibilities included designing a YAML-based configuration system for multiple LLM providers (Azure OpenAI, OpenAI and Anthropic) with Pydantic validation and fallback chains, implementing SSE streaming over LangGraph + FastAPI with custom event schemas, and building a Dynamic ReAct Tool Router with routing, execution and observation nodes. I also resolved a language bias issue in German-prompt chatbots through a "sandwich" approach with language override instructions, and handled code quality (mypy, ruff, pydoclint) and GitLab CI/CD pipeline debugging.

Management and development of automation projects using generative artificial intelligence for end clients at Accenture, where I led projects as a Tech Lead in Generative AI, working with LangChain, foundation models, Python, and APIs from Anthropic, OpenAI, Google, and other generative AI frameworks.

Lead the development of Generative AI-based products for internal use and clients. In the fintech sector