Tijmen
Denmark, United States
Summary
👤
SeniorTijmen Verhulsdonck is a Principal Software Engineer in Copenhagen with 10 years' experience building real-time, on-device perception systems for images, video and audio. He led the facial-tracking SDK at Roblox that powers live avatar animation and optimized inference across platforms by designing efficient model architectures and custom, high-performance layers. Previously he implemented AVX/SSE and fp16 weight storage optimizations in the widely used ncnn mobile inference framework and has deep expertise in squeezing latency and memory out of x86 and ARM targets. Tijmen combines hands-on low-level performance engineering with ML model design, shipping cross-platform solutions from iOS and Android to desktop, and often surfaces non-obvious, use-case specific level-of-detail tradeoffs to balance quality and throughput.
10 years of coding experience
Github Skills (16)
neural-network10
avx10
c-language10
inference10
cprogramming-language10
sse10
performance-optimization10
x86-6410
x8610
neon8
deeplearning-ai7
deep-learning7
android6
tensorflow5
vulkan5
Programming languages (5)
Github contributions (5)
Tencent/ncnn
Jun 2020 - Oct 2021
ncnn is a high-performance neural network inference framework optimized for the mobile platform
Role in this project:
Back-end Developer / Performance Engineer
Back-end Developer / Performance EngineerContributions:5 reviews, 16 commits, 17 PRs in 1 year 4 months
Contributions summary:Tijmen made several significant contributions to the `ncnn` project, a high-performance neural network inference framework. Their work focused on optimizing the framework for x86 architectures, implementing AVX/SSE instructions for various layers, including binary operations, inner product, pooling, and winograd convolution. They also added fp16 weight storage optimizations, and improved the framework's compatibility and test coverage, demonstrating a focus on both performance and code maintainability.
inferencemobilencnnneural-networkhigh-preformance
Timen/squeezenext-tensorflow
Jul 2018 - May 2022
A tensorflow implementation of squeezenext. (includes link to trained model)
Contributions:34 commits, 23 pushes, 1 branch in 3 years 11 months
tensorflow
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.