Kamal Basha

I'm an AI researcher working on underwater perception. I build physics-informed digital twins and reinforcement learning methods that help sonar AI systems work reliably when real data is scarce. My background spans heavy industry, applied machine learning, and defence research.

I am currently pursuing a Ph.D. in Computer Science and Engineering at SRM Institute of Science and Technology, advised by Dr. Athira M. Nambiar. My thesis addresses the sim-to-real gap in underwater sonar imaging. It combines physics-based sonar simulation, digital twins, generative domain translation, and RL-driven adaptive data selection so that models trained largely on synthetic data transfer to real sea conditions.

Before my Ph.D., I successfully delivered a DRDO-funded research project on AI for sonar image detection and classification. On that project I built S3Simulator, a benchmark side-scan sonar simulator and dataset (ICPR 2024), and followed it with PS3Simulator (MaCVi Workshop @ CVPR 2026). My work has appeared at venues including CVPR workshops, ICPR, ICIP, OCEANS, and JASA, and I mentor interns working on generative sonar augmentation.

Earlier, I was a Machine Learning Data Associate at Amazon. There I worked on large-scale data pipelines, annotation quality, and Python automation that reduced report generation from hours to seconds.

I started my career in Oil & Gas, supervising refinery shutdowns, piping, and maintenance for Chennai Petroleum Corporation Limited (CPCL) in India and Bahrain Petroleum Company (BAPCO) in Bahrain. That field experience shapes how I do research: I care about systems that hold up in harsh, safety-critical, real-world environments.

I hold an M.S. in Data Science from Deakin University, a PG Program in AI & ML from The University of Texas at Austin, and a B.E. in Mechanical Engineering from Anna University, Chennai.

Open to collaborations with industry and research labs in underwater and maritime autonomy, defence AI, sim-to-real learning, and digital twins. Feel free to reach out.