Sri Kishore Babu

Built an Enterprise Knowledge Intelligence & Agentic Automation Platform combining retrieval-augmented generation with multi-agent automation for employee use of large internal document repositories and multistep workflow execution. Designed and deployed a multi-tenant RAG system (LangChain, LlamaIndex, Weaviate) with hybrid dense-sparse retrieval and cross-encoder reranking, achieving query response times under two seconds and improved knowledge retrieval accuracy. Implemented multi-agent orchestration using LangGraph and AutoGen to decompose tasks into specialized sub-agents, and deployed CrewAI research/synthesis pipelines with persistent vector memory to reduce report compilation time from hours to minutes. Fine-tuned Llama 2 with LoRA/PEFT for domain Q&A, serving via FastAPI with zero-downtime rollout. Developed stateful document processing pipelines in LangGraph to classify/extract/validate hundreds of thousands of documents monthly, replacing manual review. Standardized deployme

Led an Enterprise Knowledge Intelligence & Agentic Automation platform combining retrieval-augmented generation with multi-agent automation. Designed and deployed a multi-tenant RAG system using LangChain, LlamaIndex, and Weaviate, supporting thousands of users with sub-2-second query response times. Built a hybrid dense-sparse retrieval pipeline with cross-encoder reranking and deployed it on AWS Lambda and ECS. Implemented multi-agent orchestration using LangGraph and AutoGen and built a CrewAI research/synthesis pipeline with persistent vector memory to reduce report generation from hours to minutes. Fine-tuned Llama 2 using LoRA/PEFT for domain Q&A, deployed via FastAPI with zero-downtime rollouts. Implemented a stateful document processing pipeline for large-scale classification/extraction/validation, and standardized deployments using MLflow, Docker, and GitHub Actions CI/CD.

Designed and developed a Career Coach chatbot using Python, LlamaIndex, and the OpenAI API with conversational memory and vector search for contextual relevance and scalable deployment. Built RAG systems using LangChain with OpenAI LLMs, storing embeddings in ChromaDB for real-time question answering. Engineered ML models (SVM and transformer-based approaches) for sports analytics tasks including par score estimation, and applied targeted distillation to create lightweight models that reduced inference latency and cloud inference cost. Created enterprise prompt-engineering workflows (ChatGPT-4) to improve intent detection and response relevance, and integrated pressure index scoring logic to enhance bowler recommendation engines using time-series and contextual features. Developed scalable REST APIs using FastAPI, Docker, and Kubernetes/Helm for microservice deployment and model lifecycle management. Applied generative AI and NLP for summarization, entity recognition, and natural-langu