Rishik Endurthy

Built a multi-agent LLM orchestration system using LangChain and LangGraph to improve task resolution speed in production AI workflows. Optimized NLP pipelines on GPU infrastructure to reduce inference latency and increase throughput. Created distributed data pipelines with Spark and Airflow and deployed low-latency inference services using Triton Inference Server, Docker, and Kubernetes. Integrated safety and quality components to reduce hallucination-related failures in agent pipelines.

Developed supervised machine learning models for demand forecasting and personalized recommendations, and improved their accuracy through iterative experimentation. Built NLP pipelines for classification, NER, and sentiment analysis using BERT and spaCy. Implemented real-time inference by serving models with Flask and FastAPI and improved throughput via data pipeline optimization with Spark and Airflow. Used AWS SageMaker for ML workflow and deployment and delivered business-facing dashboards for model monitoring and insights.