Roman Ageev
Staff Deep Learning Engineer (IC4), VLM at NVIDIA
Amsterdam, North Holland, Netherlands
Summary
🤩
Rockstar🎓
Top SchoolRoman Ageev is a Staff Deep Learning Engineer specializing in vision-language and multimodal LLM inference, currently optimizing production stacks at NVIDIA to cut latency, boost throughput, and reduce memory/cost on GPUs. With ~3 years of focused industry experience across startups and tech leaders (Stochastic, Yandex, Huawei, Intel, JetBrains), he brings a rare blend of research-driven modeling and hands-on systems work—quantization, FP8/INT8 tuning, TensorRT-LLM/Triton integration, and upstream fixes in vLLM and HF Transformers. He has shipped high-throughput production systems (200k rps at Yandex), built fine-tuning tooling (xTuring), and repeatedly turned SOTA ideas into robust, deployable pipelines. Based in Amsterdam, he favors simple processes for complex problems and often pairs architectural tweaks with practical profiling and automated performance gates to ensure reliability at scale.
4 years of coding experience
4 years of employment as a software developer
Master's degree, Computer Science, Master's degree, Computer Science at Higher School of EconomicsData Science, Data Science at Yandex School of Data AnalysisSaint Petersburg Lyceum 239Quantitative Finance, Quantitative Finance at Center of Mathematical FinanceBachelor's degree, Applied Mathematics and Computer Science, Bachelor's degree, Applied Mathematics and Computer Science at ITMO University
Github Skills (34)
peft10
language-model10
discord10
llama10
llm10
data-preprocessing10
deep-learning10
generative-ai10
fine-tuning10
adapter10
mistral10
lora10
quantization10
notebook9
onnxruntime9
Programming languages (2)
Github contributions (5)
stochasticai/xTuring
Mar 2023 - Dec 2023
Build, personalize and control your own LLMs. From data pre-processing to fine-tuning, xTuring provides an easy way to personalize open-source LLMs. Join our discord community: https://discord.gg/TgHXuSJEk6
Contributions:6 releases, 34 reviews, 63 PRs in 8 months
data-preprocessingdiscordfine-tuningdeep-learninggpt-2
stochasticai/x-stable-diffusion
Oct 2022 - Dec 2023
Real-time inference for Stable Diffusion - 0.88s latency. Covers AITemplate, nvFuser, TensorRT, FlashAttention. Join our Discord communty: https://discord.com/invite/TgHXuSJEk6
Contributions:3 reviews, 2 PRs, 2 pushes in 1 year 1 month
discordinferencestable-diffusiontensorrtpytorch
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.