Mathematics Undergraduate · AI Agents · Reinforcement Learning · Robotics

I build reproducible technical projects at the intersection of mathematical reasoning, intelligent agents, control, and robotics. My work emphasizes measurable results, honest limitations, and artifacts that other people can run and inspect.

A reproducible evaluation framework for testing when LLM self-correction fixes wrong answers—and when it breaks correct ones. It compares baseline, self-refine, reflexion, and tool-critic strategies on a fixed GSM8K-100 subset.

Python · LLM Agents · Evaluation · GSM8K · Reproducible Research

A warehouse AGV prototype that detects slip, estimates friction risk, and adapts speed under wet and icy conditions. In the deterministic simulation, peak slip fell from 0.95 to 0.47 with a 7.5% average-speed trade-off.

Python · Robotics · Control · Simulation · ROS 2 / Gazebo

A ten-slide technical explanation of value-iteration adaptive dynamic programming, including convergence, admissibility, finite-iteration stopping, and critic/action network implementation.

Adaptive Dynamic Programming · Value Iteration · Technical Communication

• AI agents: evaluation, self-correction, structured traces, failure review.
• Control and robotics: simulation, adaptive control, safety constraints, ROS 2 / Gazebo workflows.
• Mathematical computing: optimization, numerical methods, and reproducible experiments.
• Communication: technical reports, presentation tooling, and clear project documentation.

I am continuing to develop projects in AI-agent reliability, reinforcement learning, adaptive dynamic programming, and safe autonomous systems.