Venkata Deepak Reddy Penubaka

Developed Python-based agentic AI workflows integrating reasoning, planning, retrieval, and tool execution for complex multi-step tasks. Built distributed inference pipelines using PyTorch, vector retrieval, and memory systems to improve response accuracy and task execution success rates. Optimized serving architecture with intelligent caching, batching, and retrieval augmentation to reduce inference costs and end-to-end latency. Engineered RAG and tool-calling services supporting scalable reasoning and autonomous decision-making, and designed agent orchestration with LangGraph/LangChain/FastAPI for memory management and long-horizon planning. Implemented LLM evaluation frameworks with benchmark testing and hallucination detection, including automated quality scoring. Also built supporting microservices with Java/Spring Boot, Kafka, Redis, and PostgreSQL, deployed on Kubernetes with Docker/Helm and CI/CD. Architected distributed data pipelines with Spark/Ray/Kafka and delivered monitor

Built Python-based ML pipelines for structured and unstructured enterprise data powering predictive analytics and AI-driven decision-making used by business teams globally. Developed NLP and Generative AI solutions with Transformer models in PyTorch to improve document understanding and response accuracy for customer support/knowledge management. Delivered MLOps workflows with MLflow, Kubernetes, and automated retraining to reduce deployment cycles while improving reliability. Implemented scalable training and experimentation frameworks using TensorFlow/PyTorch/Databricks and feature engineering to improve accuracy. Created REST microservices to serve AI predictions at scale, reducing inference latency. Architected cloud-native deployments on AWS using SageMaker, S3, and EKS with monitoring and governance, and built distributed data processing pipelines with Spark/Databricks/Kafka. Deployed containerized AI services with Docker/Kubernetes and CI/CD to improve scalability and reduce inf