agents Eastside — an agentic browser you own end-to-end A desktop AI browser where the app is the browser itself (bundled Chromium) and an agent drives it with per-action consent — no browser-use, no Playwright in the path. finance Capstone Simulator — post-training MarS, and testing what a restricted agent loses Two studies built on top of the open-source MarS market simulation engine — does post-training the simulator make it more realistic, and does restricting a trading agent to model-proposed orders help or hurt execution? FBAQuant — deep hedging with RL Reinforcement-learning approach to option hedging — reward-function design for a hedging agent, grounded against the Kelly Criterion literature rather than an ad-hoc objective. Solver-Aware Deep Hedging — Paper 1 (ADI vs. Crank-Nicolson) A collaboration with Ibrahim Lanre Adedimeji (Seoul National University of Science and Technology) — benchmarking numerical PDE solvers for 2D Black-Scholes option pricing, the foundation for a planned RL-hedging paper on top. computer-vision CV-LPR — license-plate detection + recognition pipeline A two-stage computer-vision pipeline — YOLO-based plate detection feeding an LPRNet-based character-sequence recognizer — wired through a Kafka producer/consumer for streaming results.