AnilKumar Yalagala

Built a secure banking GenAI platform for enterprise knowledge search, KYC/compliance document intelligence, policy search, risk-review assistance, document summarization, internal chatbot support, and governed business workflow automation. Developed end-to-end RAG pipelines including ingestion, chunking, embeddings, vector/hybrid retrieval, reranking, prompt grounding, and source attribution with response validation and LLM evaluation (relevance, groundedness, hallucination risk, latency, cost). Implemented controlled agent workflows with tool/function calling using LangChain/LangGraph/LlamaIndex, including MCP-style patterns with tool allowlisting, schema/input validation, scoped permissions, audit logging, timeouts, and traceable execution. Exposed services via FastAPI/REST for integration with enterprise systems, and applied LLMOps/MLOps practices using MLflow, Docker, Kubernetes, and CI/CD for versioning, monitoring, and production support.

Designed scalable GenAI/LLM platforms for retail associate knowledge search, SOP/product/process retrieval, PDF/table extraction, and store-operations support. Implemented RAG architectures with hybrid search, metadata filtering, query rewriting, contextual retrieval, reranking, chunking/embedding tuning, and prompt testing to improve accuracy. Built document intelligence workflows for PDF parsing, OCR integration, document classification, and knowledge indexing. Added safeguards and human-in-the-loop for sensitive/high-impact actions, implemented fallback and model-routing strategies, and mitigated hallucinations through grounding checks, confidence thresholds, and source attribution. Containerized and deployed services with Docker on Kubernetes/EKS and delivered rapid internal prototypes using FastAPI/Streamlit, iterating with user feedback.

Developed healthcare payer ML solutions for claims analytics, member-risk scoring, care-gap identification, provider document classification, call-center text analytics, fraud/waste/abuse detection, and healthcare forecasting. Built supervised and unsupervised models for classification, regression, clustering, anomaly detection, forecasting, and recommendation-style analytics. Implemented healthcare NLP using transformer-based models (e.g., BERT) for text/document classification, entity extraction, summarization, semantic similarity, and Q&A support. Developed computer-vision support using OpenCV/CNNs. Deployed models via Azure ML and AWS SageMaker, applying MLOps with experiment tracking (MLflow), model versioning, automated training/deployment pipelines, and monitoring with governance and privacy controls.