Konstantin Gulin
Machine Learning at Cruise
Los Angeles, California, United States
Summary
🤩
Rockstar🎓
Top SchoolKonstantin Gulin is a machine learning engineer with six years of experience specializing in estimation theory, stochastic programming, and ML systems for production inference. He blends theory-driven algorithm development with pragmatic engineering, shipping performance-focused improvements in C/C++, Python, and Matlab and often bridging those languages in complex pipelines. His work at Neural Magic included concrete optimizations to the deepsparse CPU inference runtime—improving YOLO pipelines and benchmarking precision—which reflects a knack for squeezing practical gains from low-level changes. Having moved between aerospace and autonomy domains (The Aerospace Corporation, RTX, Cruise) he is motivated by the intersection of data science and the space domain and brings domain-aware modeling to real-world systems. A persistent learner with a physics and math foundation from Penn State and graduate training in data science and ML at USC, he combines rigorous analysis with an engineer’s focus on deployable, measurable impact.
5 years of coding experience
10 years of employment as a software developer
Electrical Engineering Data Science & Machine Learning, Electrical Engineering Data Science & Machine Learning at University of Southern CaliforniaBachelor’s Degree Physics & Mathematics, Bachelor’s Degree Physics & Mathematics at Penn State UniversityHigh School, High School at Huntington Beach High School
Github Skills (12)
computer-vision10
pytorch10
machine-learning10
performance-monitor10
deep-learning10
inference10
performance-analysis10
python10
onnx9
quantization9
pruning8
transformers8
Programming languages (3)
Github contributions (5)
neuralmagic/deepsparse
May 2022 - Jan 2023
Sparsity-aware deep learning inference runtime for CPUs
Role in this project:
ML Engineer
ML EngineerContributions:205 reviews, 17 commits, 25 PRs in 8 months
Contributions summary:Konstantin primarily focused on enhancing the performance and functionality of the deepsparse runtime for deep learning inference. Their contributions include optimizing YOLO object detection pipelines by switching data types and improving the benchmarking process by utilizing higher-precision timers. They also worked on exposing the `multi_label` control for YOLO pipelines, updating transformers' installation links, and resolving import issues related to transformers.
llm-inferenceruntimetensorflowsparsificationmachinelearning
neuralmagic/pytorch-image-models
Nov 2021 - Aug 2022
PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, EfficientNetV2, NFNet, Vision Transformer, MixNet, MobileNet-V3/V2, RegNet, DPN, CSPNet, and more
Contributions:4 reviews, 39 commits, 8 PRs in 9 months
mobilenetvisual-recognitionpretrained-weightsimage-classificationpretrained
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.