cv

Machine Learning Engineer specializing in perception systems for autonomous vehicles at Cruise. Based in San Francisco.

Basics

Name Jeremy Malloch
Label Machine Learning Engineer
Email jeremywebsite@malloch.ca
Url https://jeremalloch.github.io/
Summary Machine Learning Engineer specializing in perception systems for autonomous vehicles

Work

  • 2021.07 - Present
    Machine Learning Engineer
    Cruise
    Working on perception systems for autonomous vehicles
    • Unblocked Cruise's return to driverless operation by leading Unified Tracker V2 model release that resolved critical ultra-nearfield risks (largest perception risk area)
    • Contributed multiple modelling improvements to novel Unified Tracker V1 multi-modal, temporal tracking transformer model
    • Increased model flop utilization by 32%, stabilized training by analyzing model activation magnitudes, and reduced velocity flips by 20%
    • Created model attention metrics leading to 11% reduction in centroid error, 12% reduction in velocity error, and increased recall of pedestrians from 95.03% to 98.33%
  • 2019.01 - 2019.08
    Autonomous Systems Software Intern
    Apple
    Building on device ML inference infra for autonomous vehicles
    • Contributed to machine learning compiler and inference framework, using C++ and Python
    • Researched & prototyped DNN quantization & sparsification pipeline, resulting in up to 40% faster runtime of deployed models
  • 2018.05 - 2018.08
    Machine Learning Intern
    Cognitive Systems
    Developed ML models & inference tooling for WiFi based motion detectors
    • Achieved 60x improvement in neural network inference runtime by implementing im2col convolutions, vectorization and memory alignment with Eigen in C++ for ARM CPU (exceeded Tensorflow Lite baseline by 2x)
    • Researched & prototyped semi-supervised GAN data augmentation to reduce false positives

Education

Skills

Programming Languages
Python
C++
CUDA (introductory)
ML Frameworks
PyTorch
JAX
Keras

Interests

Hobbies
Hiking
Cycling
Photography
Open Source Contribution
Reading