# Python
11 posts
Building Portafel: A Research Terminal for 10,800 Instruments
Portafel covers ~10,800 stocks, ETFs, indices, commodities, FX pairs and crypto, turning company news into a personal briefing with filings, financials and price history alongside it. Here is how it is built and what was harder than expected.
The Stop Loss Was the Expensive Part: Backtesting Polymarket's 15-Minute BTC Markets
Four ways a prediction-market backtest quietly lies to you, the recorder built to avoid them, and one clear finding: adding a stop loss made every variant of the strategy worse without changing which side was right.
Teaching a Machine to Stock ATMs: Deep Reinforcement Learning for Cash Demand Forecasting
A PhD course project applying the Deep Deterministic Policy Gradient (DDPG) algorithm to ATM cash demand forecasting. The problem is framed as a continuous Markov Decision Process, evaluated against industry benchmarks on the 111-ATM NN5 dataset.
The Application of Hidden Markov Model to Detect BTC Market Regime
A research-oriented overview of how a Gaussian Mixture Hidden Markov Model can be used for BTC regime learning, with focus on theory, structure, and practical limitations.
Optimal Blackjack Strategy Through Dynamic Programming
An analysis of optimal Blackjack strategies using recursive decomposition and dealer probability analysis.
Learning Optimal Pricing with Reinforcement Learning
A technical implementation of the Actor-Critic algorithm to solve dynamic pricing. This project benchmarks RL against theoretical optima to test its effectiveness in combinatorial optimization.