Lee Mracek

EDUCATION

LINKS

COURSEWORK

SKILLS

C++ • Rust • Shell • Python
Golang • Java • Elm
C • Verilog • VHDL • LaTeX Boost • NumPy • Eigen • Tensorflow • Keras AWS • Docker • Kubernetes
gdb • Bazel • CMake • Gradle
Jenkins • Travis CI • Gerrit • Buildkite
Verilator • git FPGA interop • C++20 API Design
Memory Models & Concurrency

EXPERIENCE

• Simplified C++ interface for network communication structures to reduce indirection.
• Implemented hardware testing to identify failure modes of FPGA trading systems.
• Enhanced core C++ meta-programming framework to handle additional types of wire protocols.

• Helped design and improve graph-based execution framework for self-driving vehicles.
• Worked across the self-driving stack to provide for each team's requirements in software architecture.
• Handled complex C++ meta-programming projects using C++17 and C++20 methodologies and techniques.
• Designed and implemented deterministic simulation from logged data within the graph framework.

• Worked on research for machine learning-based clustering of radar detections.
• Developed novel tracking architecture to reduce radar tracking error.
• Collaborated with radar companies to ingest proper data on the vehicle.
• Provided advanced C++ template code to enable compile-time optimizations of tracking code.
• Decoupled compilation from requiring OS libraries, enabling build and test in minimal Docker container.
• Key role in project to move organization off of near-EOL library and software, including working with and advising many senior engineers.

• Achieved 89% classification accuracy on simulated Radar Cross-Section (RCS) data using machine learning techniques.
• Researched applications of Long-Short Term Memory networks for RCS data classification.
• Developed low-level utilities in C++ for data processing to feed into artificial intelligence methods.
• Ported RCS classification network to distributed Tensorflow on a five-machine cluster.
• Finalized proposal for ongoing deep learning initiatives, specifically Convolutional & Recurrent Neural Networks for threat classification.

• Integrated HIL simulation with Python in AutomationDesk for full test coverage.
• Wrote HIL simulation code for dSPACE using Simulink.
• Used Automation Desk to write Python unit-test cases for controls.

• Founded and am assisting a robotics team consisting of high schoolers in the community to enrich their STEM educations.
• Both founding head mentor as well as in charge of all technical work.
• Real-time programming for embedded systems and abstract Python libraries (see 'Mass' in Projects section).
• Manage back-end system including AWS, Google Apps, Confluence, Jenkins, and Gerrit.

• Managed full software stack as sole software engineer.
• Produced and tested Linux kernel modules for custom hardware.
• Worked with the CEO of Rocket EMS to evaluate hardware and software system and eliminated software as root cause after working with them for several weeks designing newest revision of a PC board.

PROJECTS

Rust framework for real-time production robotics. Work has just begun, but will include Rust compiling for bare-metal controllers, real-time optimal control, Kalman and moving horizon estimation and motion profiling. Real-time C++ framework for FIRST Robotics. Led and designed code structure and implementation. Achieves 5 ms control loops on real-time hardware with 10 μs standard deviation. Allows effective PID and full state control, Kalman filtering, DMA, and motion profiling. Built backend Python for a webapp to help De Anza College students schedule their classes.