Likith Chilakalapadi Prabhu

Developed multi-agent orchestration pipelines using PyTorch, NVIDIA NeMo, LangGraph, MCP, and MLflow to support intelligent enterprise workflows for 35,000+ employees, improving response accuracy by 22%. Built production RAG pipelines with NeMo Retriever, hybrid search, embeddings, cross-encoder reranking, and vector databases, increasing retrieval accuracy by 21%. Optimized distributed GPU inference with TensorRT-LLM, Triton Inference Server, CUDA, FlashAttention, and dynamic batching, reducing latency by 37% and lowering infrastructure costs by 28%. Designed autonomous agent workflows with planning, reasoning, memory, evaluation, and guardrails using LangGraph and MLflow tracing. Delivered cloud-native microservices and distributed execution using FastAPI, gRPC, Docker, Kubernetes, Redis, PostgreSQL, Ray, Kafka, and Kubernetes Jobs; integrated enterprise systems (SAP, Salesforce, ServiceNow, SharePoint, Confluence) with secure tool calling and governed access. Automated LLMOps and CI

Built Python-based enterprise AI pipelines using PyTorch, LangChain, RAG, prompt engineering, and MLflow for intelligent business applications used by 420,000+ users. Implemented RAG solutions with LangChain, FAISS, Pinecone, hybrid search, Hugging Face embeddings, and cross-encoder reranking, improving retrieval accuracy by 24%. Optimized LLM inference using TensorRT-LLM, Triton, CUDA, FlashAttention, and dynamic batching, reducing serving latency by 36% and lowering infrastructure costs by 32%. Designed AI workflows with memory, tool calling, hallucination detection, evaluation, and prompt versioning using LangGraph, LangSmith, DeepEval, MLflow tracing, and Hugging Face Transformers. Developed secure microservices with FastAPI, REST, gRPC, Docker, Kubernetes, Redis, PostgreSQL, and integrated secure authentication and RBAC across platforms. Implemented end-to-end MLOps/LLMOps with MLflow, Airflow, GitHub Actions, SageMaker, automated testing, and model registry. Created event-driven