Venkat Sai Jagini

Spearheaded EmergeAI and built enterprise GenAI agents on Dialogflow CX + Vertex AI Gemini, including the VZ RightSizing Agent for personalized bill-rightsizing offers. Engineered an offer-rephrasing framework with exact-figure preservation and disposition tracking for closed-loop analytics. Designed proactive resolution using LangGraph orchestration to prefetch and field-filter billing data, plus escalation risk scoring and an anticipated Q&A list to reduce handoffs. Developed production MCP servers (FastMCP/JSON-RPC 2.0) exposing playbook tools, standardized tool-calling schemas, and used feature flags for safe environment toggles. Built router-first LangGraph orchestration across retrieval, reasoning, and summarization agents, implemented RAG pipelines that improved retrieval precision and reduced hallucinations, and migrated best-performing workflows into Google Agent Development Kit (ADK). Owned regression/evaluation datasets and go/no-go release integrity via CES/Dialogflow evalu

Led development of GenAI-powered automation for financial customer service operations, including support chatbot optimization using RAG pipelines and agentic workflows (LangChain/Semantic Kernel). Built scalable LLM pipelines on AWS Bedrock and SageMaker with fine-tuned prompt strategies and embedding-based filtering to reduce hallucinations. Designed an enterprise architecture for compliant document summarization and question answering with low-latency retrieval integrated with Snowflake metadata indexing. Deployed sentiment analysis, keyword extraction, and complaint classification using Hugging Face Transformers and AWS Comprehend to improve call-center analytics. Operationalized ML and LLM services with Docker/Kubernetes, Prometheus/Grafana monitoring, MLflow experiment tracking, and FastAPI inference services on AWS EKS. Implemented secure FastAPI microservices using OAuth2/JWT, performed data mining and quality validation (Python/SQL), integrated guardrails to mitigate hallucinat

Built predictive maintenance and anomaly detection systems using time-series modeling to reduce operational costs by ~20%. Designed NodeJS microservices for real-time predictions and integrated them into backend systems. Orchestrated model refresh cycles and inference updates with Airflow and deployed models via Azure ML endpoints. Developed Azure-hosted ML APIs with Kubernetes orchestration and CI/CD pipelines (Docker/FastAPI/Node). Implemented scalable data processing with Apache Spark/Scala, built Kafka-based streaming ingestion for low-latency telemetry scoring, and created forecasting and classification/recommendation solutions using TensorFlow and PyTorch. Delivered dashboards and reporting for operations teams and optimized latency and performance for production inference.