Utkarsh Yashvardhan

M.S. Computer Science (Machine Learning), Georgia Tech | M.Sc. Mathematics & B.E. Computer Science, BITS Pilani

profile_pic.jpg

I am a mathematician and computer scientist interested in building safe and trustworthy AI on rigorous mathematical foundations. I care about how machine learning models represent what they know, how confident they should be, and how they handle data that users want kept private or removed.

I am currently completing an M.S. in Computer Science at Georgia Tech, specializing in Machine Learning (expected December 2026). Before that, I earned a dual M.Sc. in Mathematics and B.E. in Computer Science from BITS Pilani (2024). I also work as a teaching assistant for the online B.Sc. in Computer Science at BITS Pilani Digital.

Research
  • Privacy-preserving learning. For my final-year thesis at BITS Pilani, supervised by Dr Pratik Narang, I converted bokeh-rendering vision models into federated learning systems, so that user images never leave their devices. The federated models came close to their centralized baselines (details).
  • Uncertainty. At Georgia Tech, I used Bayesian inference to estimate the parameters of epidemic models from noisy data, reporting full posterior distributions rather than single estimates (details).
  • Generative AI and security. I co-authored a paper in IEEE Access (2024) analysing the capabilities and risks of generative models such as ChatGPT and DALL-E for cybersecurity (publications).
  • Scientific computing (open source). As a collaborator on SimpleTopOpt.jl, I translated published MATLAB codes for topology optimization (3D, stress-based and buckling-constrained) into Julia, and benchmarked the new implementations against the MATLAB originals for correctness and speed (Top3d.jl, TopStress.jl, TopBuck.jl).
Research interests

Machine unlearning, uncertainty and calibration, and reliability in machine learning, together with mathematical frameworks that could make these guarantees precise, such as Singular Learning Theory and Category Theory.

I am applying for Ph.D. positions in safe and trustworthy AI. If you work in this area, I would be glad to hear from you.