Rohan Parmar

Rohan Parmar is a software engineer at Mecka AI, where he works on data quality, evaluation, and reliability systems for physical AI. He previously co-founded Docula, a bootstrapped medical-AI company acquired by Mecka in 2026. His work focuses on the infrastructure required to move AI from impressive demonstrations into dependable real-world systems: trustworthy training data, failure analysis, model evaluation, provenance, long-horizon robot memory, and human-in-the-loop learning. He also researches how multimodal human signals—including egocentric video, gaze, and portable EEG—could help AI systems learn intent, uncertainty, surprise, and correction. Rohan can comment on physical AI, robotics training data, model reliability, data contamination, evaluation systems, persistent memor...