tensAI — Enterprise Agentic RAG Platform
tensAI is a sophisticated agentic RAG platform designed for enterprise scale. Built on SAP BTP, it enables seamless document ingestion, intelligent retrieval, and automated evaluation of AI responses. My work focused on architecting the core backend services, optimizing retrieval accuracy, and establishing robust evaluation frameworks to ensure high-fidelity grounded responses.
Technical highlights:
- •
Owned the Flask backend for tensAI, building retrieval workflows, document ingestion services, and integrations across SAP services.
- •
Designed a hybrid retrieval pipeline combining BM25, dense vector search, and reranker, improving top-3 retrieval relevance by 38%.
- •
Built multi-format document parsers for PDF, PPTX, DOCX, and Excel, significantly improving grounding reliability in RAG workflows.
- •
Developed a RAGAS-based evaluation framework for automated hallucination monitoring, reducing evaluation turnaround time by 60%.