Rahul Padinhappurath Bharathan

Researching and validating Conformal Prediction methods to make Al safe for aerospace users. I am treating this as a discovery project to establish valid confidence intervals for future production systems. Benchmarking various calibration algorithms to rigorously detect and reduce hallucinations, ensuring the system knows when to "abstain"-a critical feature for maintaining user trust in enterprise tools.

Core contributor to 'GenAir', architecting Agentic workflows that move beyond simple chat to solving multi-step business problems. I helped translate these initial experiments into production-ready data products. Established evaluation standards for LLM deployment and shared. MLOps best practices across the team, helping other developers adopt standardized testing for AI models. Built scalable evaluation pipelines using Jenkins and Docker to automate model testing.