Projects

I’m interested in quantitative research that enables us to understand the properties of complex systems including the Finance using the principle of physics. Here, I share the research projects pertianing to Machine Learning and Statistical Physics its application in finance, complex systems and quantitative image-processing.

Here are some of my notable projects:

• Credit Risk Analytics & Real-Time Scoring (Lending Club Case Study) - Built an end-to-end pipeline for feature engineering, model training, and a live scoring web-app. [Data Science in Finance]

• Graph-Based Pattern & Anomaly Detection from Images - Converted images to networks to detect endpoints/junctions; framed as graph analytics for fraud/risk signal discovery. [Quantitative Image Processing]

• Interactive Analytics App for High-Dimensional Data (Chemical Fingerprint) - Developed a GUI to slice, filter, and visualize large matrices—analogous to building BI tools for portfolio/credit dashboards. [Soft-Matter Physics/Analytical Chemistry Analysis]

• Optimization-Driven Signal Reconstruction (Phase retreival algorithm) - Implemented inverse-problem methods to recover missing/noisy signals for optical holography; transferable to data imputation and time-series smoothing in finance. [Quantitative Image Processing]

• Monte Carlo Simulation (Brownian motion) - Built stochastic simulators (Langevin/agent models) to study ecological problem -> transfareable to generate scenarios applicable to VaR, liquidity, and credit stress testing. [Statistical Physics and Thermodynamics/Machine Learning]

• Dynamic Phase Transitions in Non-Equilibrium Systems (Nonequillibrium Physics) Converted noisy microscopy into quantitative signals via a MATLAB→Python pipeline (denoising, segmentation, feature extraction); modeled dynamics, detected jamming transitions; published in PNAS Nexus. Code/Data · [Statistical Physics and Thermodynamics]
  Media coverage: ScienceDaily · Phys.org · Bioengineer.org · University of Tokyo