applied research
Applied research portfolio.
AREAS OF EXPERTISE
- Forecasting & Predictive Modeling: Specialized in simulating and predicting the behavior of complex, nonlinear systems using traditional (ARIMA, XGBoost, RNN) and innovative AI-driven approaches that combine neural operators and physical modeling.
- Uncertainty Modeling: Skilled in Bayesian inference, bootstrapping, quantile regression, and censored data techniques to ensure robust risk assessment and probabilistic forecasting in high-stakes environments.
- Advanced Machine Learning & Data Analysis: Proficient in deep neural networks, transformers, and physical neural networks, with extensive practical expertise in Python and CUDA—optimizing models for speed and accuracy.
- High-Performance Computing & Simulation: Extensive experience with GPU-accelerated computing (CUDA/MPI) to reduce simulation times significantly, which is directly applicable to real-time forecasting and large-scale quantitative analyses.
- Quantitative Research & Mathematical Modeling: Deep background in advanced mathematics to formulate robust models that capture the dynamics of chaotic systems, an expertise that transfers well to quantitative finance and risk management.
SELECTED PROJECTS
- AI-Driven Forecasting Algorithms: Developed a PyTorch-based forecasting algorithm using neural operators to accelerate simulations (up to 25x faster), showcasing techniques that can be tailored to forecast hard-to-predict systems.
- GPU-Accelerated Simulations: Pioneered GPU solvers using CUDA to reduce simulation runtimes by 3x in complex environments, demonstrating the capability to take advanced simulation techniques from research and deploy them efficiently in high-performance financial modeling and uncertainty quantification.
- High-Performance Bayesian Inference Engine: Combined deep learning with Markov chain Monte Carlo methods to swiftly generate and fit models that can be directly translated into probabilistic forecasting and risk management.
- Uncertainty Quantification in Limited Information Environments: Developed regression techniques for censored, sparsely sampled data and applied bootstrapping to robustly quantify prediction uncertainty.
- Feature Selection from High-Dimensional Data: Led research published in Science that distilled dozens of variables into two key predictors, isolating the core drivers of system behavior—a crucial capability for alpha generation in quantitative finance.
INDUSTRY APPLICATIONS
- Market and demand forecasting: Expertise from black hole dynamics can predict behavior in similarly complex financial markets and transportation networks.
- Risk Assessment: Tail risks and rare events.
- Alpha Discovery: Repurpose signal extraction methods from noisy astronomical data to identify hidden market inefficiencies.
- Computational Advantage: Accelerate backtesting and simulations by orders of magnitude.
PUBLICATIONS AND GRANTS
- Authored 100+ publications in top journals (e.g. Science) with 10,000+ citations.
- Obtained US$800k in grants as PI from NASA and FAPESP.