Shiva Kaul. I'm a Ph.D student in the Computer Science Department at Carnegie Mellon. I work on trustworthy machine learning. Like most people, I've been pretty amazed by the recent advances in AI and ML; like most researchers, I'm kind of wistful for the clarity and understanding we had in traditional machine learning, and a bit concerned about how a lack of such understanding could potentially lead to bad, harmful decisions. But I'm optimistic: In my opinion, modern learning is not inherently inscrutable, and classical learning isn't merely relegated to the dustbin of history. I think by nontrivially combining classical models with modern ones, we can achieve the best of both worlds. For example, we can replace the guts of RNNs or Transformers with linear systems, making them both faster and more mathematically tractable. Alternatively, we can wrap foundation models (such as LLMs) in Gaussian processes and conformal prediction, allowing them to reliably answer important causal questions. I've always been interested in applications to healthcare and wellness. With these new techniques, I'm currently reexamining the foundations of evidence-based medicine.

My advisor is the razor-sharp, incredibly-patient Geoff Gordon. Earlier, I earned an M.S. under Mahadev Satyanarayanan on the topic of human-in-the-loop machine learning.

Publications, reports, and working papers

Selected recent talks

Contact me