I build machine-learning and optimization tools for forecasting and quantitative finance
15 pull requests merged into libraries used across ML and finance, most of them fixing silent numerical errors:
- Apple MLX: reductions over very large arrays (32-bit index overflow on CPU and GPU) (#4549)
- Amazon Chronos: series lengths for categorical item ids in
predict_df(#537) - CVXPY (5): MIP warm start for SCIP (#3483),
geo_meanon any shape (#3493), tuple axes instd, axis behavior ofcvar, axis checks for matrix norms - skfolio (3): VaR, EVaR and sample-weight fixes in the risk measures
- NautilusTrader (2): wrong default moving averages in trading indicators
- Also Qdrant, NumPyro and Qiskit Experiments
In review: Microsoft FLAML (approved), the NVIDIA cuOpt interface in CVXPY, Unit8 Darts and Nixtla (conformal prediction intervals), QuantEcon.
- WorldQuant BRAIN: #2 in Italy, Gold Level, 19 submitted alphas
- Kaggle: silver medal, 76th of 4,186 teams, AI Agent Security (OpenAI × Google × IEEE)
- MQFS: open, reproducible quantitative-finance papers, e.g. market regimes from correlation structure, electricity risk for AI data centres, constraint-preserving QAOA on IBM quantum hardware
- Qiskit Advocate (IBM Quantum) and solver of Jane Street's monthly puzzles
Open to ML engineering and quant, swe, data, or anything similar, including remote and part-time contract work or project-contract. LinkedIn

