Work

Custom AI/ML Implementations

Python
TensorFlow
Genetic Algorithms
Bayesian Optimization

Custom-built neural networks, genetic algorithms, Bayesian optimization, and stochastic models in Python. Built at Xcapit alongside a team of 5 PhDs.

Neural network visualization

Business Case

At Xcapit, we needed AI models that went beyond off-the-shelf solutions. Our portfolio optimization and trading strategies required custom implementations with full control over the math. We built everything from scratch, before ChatGPT existed.

Impact

  • Team of 5 PhDs (mathematics, physics, computer science) building custom AI
  • Pre-GPT era: all models designed and implemented from scratch
  • Powered the trading engine that achieved 400%+ annual returns
  • UNICEF and Ethereum Foundation funded the broader platform

What’s Inside

ModulePurpose
GeneticAlgorithm.pyCustom genetic algorithm for combinatorial optimization
ModelOptimization.pyHyperparameter tuning strategies
ModelValidation.pyRobust cross-validation and evaluation
simulations.pyMonte Carlo simulations and stochastic calculus
LSTM modelsTime-series prediction for price forecasting

View source code on GitHub →