Hey all, I'm Arun. I just wrapped a CS PhD at the University of Minnesota on physics-informed ML for spatial AI, basically getting models to know when they're wrong.
Most of what this community cares about (agent reliability, evals, observability, failure handling) is what my research has been about, just from the spatial side: anomaly detection on GPS and maritime trajectories, evidential uncertainty as an eval signal, and physics-informed diffusion. Lately I've been building multi-agent spatiotemporal systems too.
I'll be at Observe tomorrow and would love to meet folks in person. I'm also on the market for research and applied scientist roles, so if you work on reliability or evals (or hire for it), say hi.
What are you all building? Always happy to trade notes on making AI trustworthy enough to actually ship.
