- Designed and operated end-to-end automated trading system for US equities using Python and q/kdb+- Built high-performance data processing pipeline combining q/kdb+ with custom C++ plugins for market data handling- Developed deep learning prediction models using PyTorch and Keras, incorporating custom architecture and using CUDA-accelerated computing, with careful CPU/GPU load balancing- Implemented asynchronous order management system with adaptive execution logic, using asyncio to handle concurrent order flows and dynamic price adjustments based on observed market impact- Engineered custom auction price computation system using tick-level data to improve model training accuracy- Built full infrastructure stack across local and cloud Linux/Windows environments, including automated data collection, strategy execution, and risk management

- Designed and implemented recommendation systems with collaborative filtering from multiple data sources- Created and published methodology for improving statistical power of A/B tests using Fisher's method

- Built large-scale ML systems using Spark/XGBoost on enterprise Sun Grid Engine, processing distributed workloads- Drove code reusability initiatives across research teams, standardizing ML components and workflows

Worked alongside traders on high-performance time series analysis system, implementing mathematically optimized sliding-window regression algorithms in q/kdb+ on cryptocurrency data

Developed and taught ML curriculum including Deep Belief Networks, Gradient Boosting, and Information Theory

Developed optimization approaches for resource allocation using Hadoop/Spark ecosystem

- Designed and implemented company's core music composition engine, re-writing it from scratch, using RNNs and rule-based AI, with Python/C++ production implementation- Consistently fixed critical production bugs in timeframes of minutes to hours while maintaining testing standards - Managed junior AI researcher and drove research direction, contributing to company's TechCrunch Disrupt 2015 win

- Trading algorithm analysis and signal research utilizing tick and level 2 market data, implementing high-speed custom machine learning algorithms using q/kdb+, Python, R, C++, Java and Fortran in delta-one strategies- Spearheaded development of shared q/kdb+ libraries by proactively identifying common patterns across team workflows; actively collaborated with team members to ensure usability and adoption, while introducing Kanban to improve code sharing and development practices

- Built analytics platform bridging siloed departmental data, establishing common metrics definitions and automated reporting infrastructure using R and bash scripting- Successfully mediated between management, marketing, and technical teams to align data interpretation and business objectives, culminating in VC presentations and successful funding rounds

Covered unsupervised learning, including EM, variational methods, expectation propagation, etc.

Covered probability theory, HMMs, reinforcement learning, mathematical logic