AI Systems Architect • Multi-Agent Systems • Deep Reinforcement Learning
Building AI-powered field service management at Exoserva
Multi-Agent Systems → 4 AI agents coordinating via stigmergy (80% token reduction)
Deep RL Pipelines → PPO, DQN, Rainbow from scratch in PyTorch
ML Infrastructure → AutoML, XGBoost pipelines, SHAP interpretability
DDD Architecture → Event-sourced domains for financial analytics
| Project | What It Does | Tech |
|---|---|---|
| autonomous-agents | 4 AI agents collaborate like ant colonies | Claude API, Stigmergy |
| ml-ppo | PPO from scratch with GAE | PyTorch |
| ml-dqn | Rainbow DQN: Double, Dueling, PER, Noisy | PyTorch |
| ml-xgboost | XGBoost + SHAP interpretability | Python, FastAPI |
| ml-common | 768-dim state vectors for trading | NumPy, Numba |
More ML Projects
| Project | Description |
|---|---|
| ml-automl-pipeline | Automated ML with Optuna |
| ml-volatility-forecasting | GARCH models for time series |
| ml-anomaly-detection | Isolation Forest, LOF, Autoencoders |
| ml-explainable-ai | SHAP, LIME interpretability |
| ml-meta-learning | MAML, Reptile, ProtoNet |
Core platform repos are private — commercial product in active development.