Vincent Brisse

Led the design and deployment of enterprise-scale ML solutions across multiple industries. Headed predictive maintenance projects for oil & gas clients, an energy forecasting and management application for a major energy company, a low-latency fraud detection system for a financial institution, and an inventory optimization project for a retailer. Contributed to core product development and GenAI/LLM capabilities, including distributed training with Accelerate/DeepSpeed/FSDP on LLaMA-style models, bootstrapping data, and DPO for output safety. Built a multimodal RAG evaluation framework with custom retrievers (ColBERT, FAISS, Chroma) and a contextual QA agent with tool-calling and memory, plus a modular query-to-spec tool-calling pipeline for end-to-end performance.

Designed and deployed enterprise-scale machine learning solutions for clients across multiple industries. Led development of predictive maintenance applications for three oil and gas customers, an energy forecast and management application for a leader in the energy sector, a low-latency fraud detection system for a major financial institution, and an inventory optimization application for a retailer. Contributed to core development of C3.ai product suite and GenAI/LLM capabilities, including distributed training with Accelerate/DeepSpeed, multimodal RAG evaluation framework, and a query-to-spec tool-calling pipeline.

Designed and deployed enterprise-scale machine learning solutions for clients across multiple industries. Led predictive maintenance applications in the oil and gas industry for three customers. Led the implementation of an energy forecast and management application for a leader in the energy industry. Led the development of a low-latency fraud detection system for a major financial institution. Worked on an inventory optimization application in the retail industry. Contributed to core product suite and GenAI/LLM capabilities: distributed training and alignment framework using Accelerate, DeepSpeed, FSDP for LLaMA‑style models; bootstrapping data; applied DPO for output safety. Designed a multimodal RAG evaluation framework; built custom retrievers (ColBERT, FAISS, Chroma) and a contextual QA agent with tool-calling and memory for dynamic few-shot prompting. Developed a query-to-spec tool-calling pipeline: decomposed complex natural queries into structured queries (DB access, filters