Consulting Case Studies

Applied mathematics, data engineering, and predictive modeling projects delivered by our specialized student teams to industry partners.

Financial Time Series & Risk Client: Inspiration-Q
Completed

Covariance Matrix Estimation in Financial Time Series

Benchmarked out-of-sample covariance forecasting methods for portfolio optimization. Evaluated sample covariance, adaptive EWMA, Ledoit-Wolf shrinkage, and random-matrix denoising across sliding windows.

Technologies & Tools
Python NumPy Riskfolio-Lib Scikit-Learn
Key Deliverables
  • → Systematic backtesting framework evaluating in-sample vs. out-of-sample forecasting.
  • → Adaptive EWMA model with dynamic half-life tuning to handle market volatility shifts.
  • → White Paper and Jupyter Notebook suite detailing optimal estimation methods.
Market Microstructure & Trading Client: Iqana
Completed

Intraday Liquidity Modeling in Crypto-Assets

Analyzed high-frequency order book dynamics across crypto pairs (BTC/USDT, ETH/USDT). Built 1-minute aggregation pipelines and computed a composite Liquidity Score to evaluate TWAP and VWAP execution.

Technologies & Tools
Python Pandas Order Book Telemetry Microstructure Analytics
Key Deliverables
  • → Data engineering pipeline collecting Level 2 order books and trades with normalized timestamps.
  • → Composite Liquidity Score synthesizing relative spreads, market depth, and slippage.
  • → Execution benchmark relating liquidity conditions to algorithmic order performance.
Healthcare IoT & Machine Learning Client: YASYT Robotics
Completed

Prediction and Detection of Risk Events in Older Adults Using Multisensor Data

Developed risk detection models using multisensor IoT telemetry for YASYT Robotics. Augmented anonymized in-home alerts with targeted synthetic profiles to train models separating true emergencies from false alarms.

Technologies & Tools
Python Scikit-Learn IoT Multisensor Telemetry Synthetic Data Augmentation
Key Deliverables
  • → Exploratory analysis on real in-home alert data, uncovering key behavioral risk patterns.
  • → Synthetic data generation simulating diverse mobility profiles to resolve class imbalance.
  • → Supervised classification models distinguishing genuine risk events from false alarms.

Have a Complex Mathematical Challenge?

Our student teams tackle complex machine learning, operations research, data modeling, and algorithmic development.

Propose a Project Challenge