UAI 2026
A non-convex but invex loss (LEARN) that tolerates an unknown number of corrupted rounds, with tight dynamic regret guarantees and no Lipschitz assumption.
Anand Krishna
Lead AI Scientist at PoiroBengaluru, India
I lead AI/ML at Poiro, working on generative AI and agentic systems for creative production. My research background is in game theory, online learning, and randomized algorithms, where I studied decision-making and learning under uncertainty. My current focus is on building reliable AI systems that can reason, use tools, and carry out creative workflows within real-world production systems.
About
Before Poiro I was a research consultant in Walmart’s AdTech team, and a Postdoctoral Research Fellow at the National University of Singapore, working with Prof. Vincent Y. F. Tan on outlier-robust optimization and multi-armed bandits.
I obtained my Ph.D. as a Prime Minister's Research Fellow (PMRF) in the Department of Computer Science and Automation at the Indian Institute of Science, where I was fortunate to be advised by Prof. Y. Narahari and Assoc. Prof. Siddharth Barman.
Interests
- Game Theory
- Online Learning
- Randomized Algorithms
- Reinforcement Learning
Education
- Direct Ph.D., Intelligent Systems
- B.Tech, Computer Science & Engineering
Selected work
All publicationsIEEE Transactions on Information Theory
Recovers a signal from magnitude-only measurements under arbitrary corruption, with nearly linear sample complexity and no spectral initialization.
-Mean Regret for Stochastic Bandits
ConferenceAAAI 2025
A single UCB-based algorithm that achieves -mean regret bounds across the whole fairness–efficiency spectrum, recovering average and Nash regret as special cases.
News
“Dynamic Regret in Outlier-Oblivious Online Optimization using Nonconvex Robust Losses” accepted at UAI 2026.
“A Sample Efficient Alternating Minimization-based Algorithm for Robust Phase Retrieval” published in IEEE Transactions on Information Theory.
Joined Poiro as Lead AI Scientist, leading Data Science & AI.
“-Mean Regret for Stochastic Bandits” accepted at AAAI 2025.
“Sublinear Approximation Algorithm for Nash Social Welfare with XOS Valuations” accepted at ITCS 2024.
Teaching
Probability Models
Computational Methods of Optimization
Game Theory
Randomized Algorithms
Linear Algebra and Probability
Get in touch
Happy to talk about fair division, online learning, or anything adjacent — email is best.
anandkrishna1995@live.com