At Lazard, I am part of the team that developed LazardGPT, an internal ChatGPT platform, by integrating various AI models such as Azure OpenAI, Anthropic, and LLaMA. I enhanced the platform by introducing interactive assistants, including a code interpreter for data analysis on uploaded files and retrieval assistants to fetch relevant information from internal documents, significantly improving user engagement. In the next phase, we focused on developing APIs to integrate assistants with tools for retrieving documents. Leveraging a large repository of company documents, we ingested them into a Milvus vector database to enable semantic search and retrieval. I designed and developed FastAPI endpoints to process user prompts, query the database, and return relevant documents in real-time. Additionally, I set up end-to-end ML pipelines to fine-tune and deploy Large Language Models
(LLMs) in production environments using Azure Kubernetes Service (AKS) and GPU-accelerated infrastructure.
