AI Safety & Systems Researcher

I'm GiGi Koneti, an AI safety and systems researcher from Bengaluru, focused on evaluating, verifying, and stress-testing generative and physical AI models.

I work on building benchmarks for physical consistency (mass and momentum conservation), calibrating probabilistic forecasts, and auditing GNN and diffusion architectures for safety failures.

I actively contribute to PyTorch/ao, Hugging Face, and FOSSEE/IIT Bombay, and run a research podcast interviewing AI researchers on the latest advancements.