Project History
Retrieval-Augmented Generation (RAG) System
Led the transition of a student thesis into a mission-critical, production-grade RAG system deployed on Azure and Kubernetes. Oversaw architectural design, implementation, and scaling of the platform for enterprise use. The system combines vector-based semantic search with large language models to enable accurate, context-aware responses.
AI-Powered HVAC Optimization
Designed and implemented data science infrastructure capable of ingesting and processing high-velocity, real-time sensor data streams from hundreds of buildings. Built machine learning algorithms to detect inefficiencies in HVAC systems and continuously recommend operational improvements.
Machine Learning Forecasting System
Introduced machine learning to replace a manual model generation process for macroeconomic forecasting. Designed models grounded in economic principles but capable of unbiased walk-forward testing. The new models delivered better live performance and enabled rapid diversification across markets. Tools and technologies used included Pandas, TensorFlow, scikit-learn, NumPy, and Ray. Led the development of the machine learning framework and built a team of quantitative researchers to bring the system into production.
Research & Trading Infrastructure
Rebuilt the company’s data science and trading platform from Access/Excel to a modern, modular Python-based system. Deployed across two data centers and an on-premises Linux compute cluster. Core technologies included S3, Redis, MongoDB, PostgreSQL, and Jupyter. Developed a Research and Trading SDK to support research, model development, automated retraining, and trading signal generation for 400+ machine learning models.
Reconnaissance Vehicle Prototype
Developed embedded control software for a reconnaissance vehicle, routing messages and sensor data across subsystems. Built the user interface for operating optical and range-finding instruments used in field operations.
Combat Simulation Systems
Produced reports and prototypes on advanced algorithms for military simulation systems. Topics included genetic algorithms for strategy optimization, novel routing algorithms for dynamic environments, and simulation-based evaluation of incremental system improvements.
Stock Exchange Infrastructure
Architected and developed the full technical stack for a Swiss stock exchange. Work included the matching engine, market data APIs (C/C++), compliance automation, and secure VPN-based member access, approved by regulatory authorities. Designed the system’s infrastructure, security, and network topology.
Variance Swap Visualization Tool
Built an interactive variance swap pricing tool that interpolated sparse quotes across counterparties and time horizons. Designed with a focus on clarity and usability for traders, the interface allowed for visual validation of market data and theoretical model outputs.
Volatility Hedging Strategy
Extended an intraday delta-neutral options strategy with a real-time volatility surface fitter and dynamic strategy adaptation. Innovations included speculative gamma-based order placing and adaptive behavior in response to market regime changes. Enhancements significantly improved performance, transforming it into the fund’s most profitable strategy.
Modular Trading Platform
Led the architecture and implementation of a distributed trading infrastructure to replace a monolithic system. Built modules for real-time risk management, historical backtesting, order routing (in Erlang), and high-frequency data integration via fiber links to brokers and exchanges.
Pedestrian Simulation Framework
Developed a modular C++ simulation engine for modeling pedestrian movement, based on PhD research. Widely adopted in academia, motion picture production, and architectural visualization.