AI Data Scientist

Detroit, MI (Hybrid/Onsite Tue, Wed, Thu) W2 contract role 6 Months then eligible for Contract renewal Role Overview The Advanced Analytics Hub team is looking to bring on board an Expert AI Data Scientist for an AI project as a Contractor. The objective of this project is to build an intelligent, agentic AI solution that provides material recommendations from SRM Material Catalogs at the point of purchase requisition, leveraging RAG and LLM-based capabilities. Key Requirements: • End-to-End Agentic RAG System Design – Proven experience designing and deploying production-grade RAG systems, including embeddings, vector search, and agent orchestration for multi-step reasoning workflows. • LLM & GenAI Integration at Scale (with Agent Frameworks) – Hands-on expertise integrating LLMs into enterprise applications, including prompt engineering, tool usage, and experience with frameworks such as LangGraph/LangChain, Semantic Kernel, or AutoGen. • Retrieval Quality, Evaluation & Optimization – Strong background in evaluation frameworks (precision/recall, grounding accuracy, hallucination detection) and optimization techniques (chunking, re-ranking, hybrid search). • MLOps & Productionalization – Experience deploying AI solutions at scale with CI/CD pipelines, model lifecycle management, monitoring, and cloud environments (Azure preferred). • Strong ML & Statistical Foundation – Deep expertise in Python, ML/statistics, and experimentation with a focus on rigorous validation of model performance and business impact. • Systems Thinking & Enterprise Integration – Ability to architect and integrate AI solutions within enterprise ecosystems (preferably SAP/SRM or similar procurement workflows). • Vector Databases & Retrieval Infrastructure – Hands-on experience with vector databases (e.g., Azure AI Search, Pinecone, FAISS) and optimization for real-time use cases. • Core Engineering Best Practices – Strong proficiency in Python, Git, API development, and modern software engineering practices including CI/CD for ML systems. • Experience running local models – we run local models on high compute servers on prem (500 GB RAM, 4 L40s GPUs, 64 cpu running on RHEL 9.7) before deploying the solution on cloud platform to save us the cloud cost during development