David Durst Email & Phone Number

San Mateo, California, United States

Stanford, California

Efficiently Imitating Human Movement in Counter-Strike- Analyzed human behavior traces and trained model to imitate humans- Trained transformer-based learned movement controller in AI performance constraints of a commercial video game- Curated 123-hour dataset of expert human movement- Designed and executed user study demonstrating learned movement model best imitates humans- Created quantitative metrics evaluating similarity to human behavior distribution- Managed online community of subject matter experts for user study and general feedbackAetherling: Type-Directed Scheduling of Streaming Accelerators- Created languages: express image processing applications that compile to statically scheduled, streaming hardware accelerators- Developed space-time type system: express trade-offs between throughput and resource utilization- Implemented auto-scheduling compiler: trade-off throughput and resources while preserving semantics with type-directed rewrite rules- Generated FPGA designs: use fewer resources than designs created by comparable systems

- Developed software for behavioral anti-cheat feature and deployed in production Call of Duty game- Modeled player churn using survival analysis techniques

- Prototyped Halide-to-FPGA compilation toolchain, modeled performance on FPGAs, CPUs, GPUs, and DSPs- Predicted that DSPs offer better performance per watt for target applications than FPGAs

Greater New York City Area

- Designed philosophy and led development of TopNotch big data quality control system, managed team of developers in US and India- Architected and implemented frontend and backend for interactive big data visualization system- Worked with mortgage-backed security (MBS) portfolio managers to model leading indicators of asset performance using loan level data sets- Researched how to use machine learning techniques to value financial assets

Greater New York City Area

- Researched visualization techniques for integrating big data into investment decisions- Implemented software package that identifies market patterns- Worked in 10 person team to create a mock, long-only Russell 1000 portfolio

Westport, Ct

- Optimized economic models to enable investment teams to react to market changes- Created profiler that identifies areas of low performance

Greater New York City Area

- Created and led HIPPA compliant testing for software that identifies unused files- Optimized software development cycle by decreasing compilation wait times up to 83%

Doctor Of Philosophy - Phd, Computer Science

Bachelor Of Science In Engineering, Summa Cum Laude, Computer Science

Activities and Societies: Graduating Class Committee, Class of 2015; Club Table Tennis, Vice President; The Princeton Charter Club.