Learn. Understand. Distribute.
My name means "vast knowledge." I am spending my life living up to it, learning across domains spanning from physics to the mind. My mission is to maximize knowledge through sustainably growing the scientific community. To achieve my goal, I will learn topics efficiently, understand them deeply, and distribute them strategically.
The World is not a Taylor Series
I started in Physics, inspired by von Neumann and Wolfram, hoping to describe the world with perfect equations. But I realized the world is not a Taylor Series. Real systems are messy that cannot be described with clean formulas.
I switched to AI to deal with that reality. While physics often ignores the noise, statistical learning treats uncertainty as information. This lets me build useful models even when exact answers don't exist.
The Positive-sum Game
Learning in isolation has a limit. It creates a bottleneck where insights stop at the individual. To truly maximize knowledge, I had to go further. Explaining a complex concept in a clear way is my approach to achieve deep understanding. This makes distributing the final step. By lowering the barrier for others, the community grows.
Bachelor's Thesis | First Author Paper
The Temporal Frequency Network
In signal processing, you cannot perfectly localize a signal in both time and frequency simultaneously. Standard models force a choice.
I engineered a heterogeneous architecture that fuses Causal Convolution with Spectral Convolution. This achieves a superior representation of reactor physics under strictly fixed compute budgets, running legacy simulations in 12ms.
Master's Thesis | Theoretical Statistics
Uncertainty Quantification for ROC Curves
Speed is nothing without trust. Existing methods for measuring AI performance (Bootstrap, Bayesian) often suffer from bias or subjective priors.
I derived a method with finite-sample guarantees, distribution-free, computationally efficient, and theoretically rigorous to solve this problem.