Driven by intuition from the human brain, I aim to build reliable AI systems capable of operating in complex, unstructured environments. My work bridges the gap between theoretical first principles and practical implementation—ranging from solving complex mathematical Problems via Reinforcement Learning and PINNs to the ground-up development of reasoning models.
My engineering philosophy centers on "from-scratch" implementations. I build core architectures—from Qwen-compatible LLMs and custom tokenizers to DeepSeek-style GRPO reinforcement learning loops—entirely in PyTorch. Furthermore, I connect the theoretical principles of Stochastic Calculus (Wiener processes, Ito's formula) to continuous-time generative AI, bridging pure mathematics with the mechanics of diffusion models.
