This project addresses the real-world portfolio optimization problem, going beyond classical mean-variance models. Actual portfolio construction involves discrete investment decisions, transaction costs, and monitoring constraints, making the problem a Mixed-Integer Optimization (MIO) challenge that is computationally intractable at scale

Yong Zhang, Jiahao Li, Xingyu Yang, and Hong Lin. "Aggregating exponential gradient expert advice for online portfolio selection under transaction costs" Journal of the Operational Research Society, , 2023, 74(8): 1940-1953.

End-to-End Python implementation of Devanathan et al.'s (2026) ADMM-based distributed optimization for institutional market impact mitigation. Features 3/2-power transaction cost modeling, proximal operator calculus, VAR(1) alpha generation, and 25-year walk-forward validation, via backtesting, across 434 assets. 

Systematic BIST equity portfolio toolkit - band rebalancing, transaction-cost modeling (Abdi-Ranaldo / Kyle), volatility calibration. Python, tested, CI-gated.

Transaction-cost-aware portfolio optimization using the Gârleanu–Pedersen framework, VAR-based signal dynamics, and PPO reinforcement learning on factor-driven portfolios.

Intraday mean‑reversion trading strategy with realistic execution costs, turnover caps, and walk‑forward backtesting.

Early economic framework for YIELD, a project at crossroads between DeFi and my economics work. This presents an economic analysis of optimal capital allocation across DeFi lending protocols, specifically for populations in emerging countries (with specific constraints and lower capital available).

ML regime detection for adaptive offensive/defensive factor allocation in US equities (1980–2024) — reproducible pipeline & cost-aware backtests.

PyTorch Dueling-DDPG for American put option hedging under transaction costs (0%/3%), benchmarked vs delta hedging. Project: Oct 2024–Oct 2025.

Backtest engine with realistic transaction-cost model (high-turnover wins gross, breaks even net).

Research-only cost-viability study showing why Binance–Bybit perpetual funding spread arbitrage failed under realistic non-VIP retail execution assumptions.