Looking for a job

Grigorii Alekseev

@grin606

Message

Principal GPU Systems Engineer | GPU Architecture & Performance | CUDA

Georgia

Message

What I'm looking for

I’m looking for a senior, staff, or principal-level role with ownership of GPU systems end to end: architecture, execution, memory management, profiling, integration, and developer-facing interfaces. I’m strongest where a company needs to build or substantially redesign a GPU solution, not only optimize isolated kernels. I prefer remote-first teams and broad system-level responsibility.

I specialize in taking ownership of GPU systems end-to-end — designing and building them from scratch, or taking existing GPU stacks and improving them across architecture, interfaces, algorithms, and low-level performance. I work across the full stack: system architecture, CPU/GPU boundaries, execution and memory models, algorithms, CUDA kernels, profiling, and integration with the rest of the product. Over the past 10+ years, I've done this across computer vision and AI inference, automotive and radar, GPU databases, and cryptographic systems.

Experience

Work history, roles, and key accomplishments

GPU Expert

Axiom / OpenVM

Dec 2024 - Sep 2026 (1 year 9 months)

Owned GPU architecture from first design through integration with the Rust proving system, shipping an open-source GPU prover. Designed an asynchronous virtual-memory-based manager reducing peak GPU memory use by 30% and delivered a Halo2 GPU implementation 20x faster than the original.

CUDA GPU Architecture Rust Memory Management Performance Optimization Halo 2 ZK Proving Mentorship

GPU Expert Consultant

Succinct

Jul 2024 - Oct 2024 (3 months)

Profiled GPU implementations across ZK proving workloads, optimized bottlenecks, and implemented CUDA kernels, accelerating a target operation by approximately 200x. Demonstrated multi-stream execution architecture and developed arithmetic kernels for experimental cryptographic fields.

CUDA Performance Profiling ZK Proving Multi Stream Delivery Arithmetic Computations Mersenne Binius

GPU Performance Engineer

Perplexity

Sep 2023 - Jun 2024 (9 months)

Collaborated with NVIDIA's TensorRT-LLM team to identify bottlenecks and shape optimization plans. Adapted neural-network models in Python and implemented fused kernels for FlashAttention.

NVIDIA TensorRT LLM CUDA Flash Attention Python Neural Networks Performance Optimization

GPU Architect / Senior CUDA & C++ Developer

SQream

Jun 2021 - Mar 2024 (2 years 9 months)

Built a GPU database project with a six-engineer team, designing execution pipeline, server architecture, and custom CUDA algorithms, achieving up to 12x faster queries. Reviewed GPU architecture across the company and developed proofs of concept for new GPU approaches.

CUDA C++GPU Database Architecture Performance Optimization Proof of Concept

CUDA Software Developer

Intel

Jan 2019 - Mar 2021 (2 years 2 months)

Developed low-latency CUDA computation for augmented-reality glasses on NVIDIA Xavier, translating mathematical algorithms from C++/MATLAB to real-time GPU implementation. Wrote a custom Fourier transform for an autonomous-vehicle radar prototype that outperformed NVIDIA cuFFT.

CUDA C++Matlab NVIDIA Xavier FFT Real time processing

Senior CUDA Developer

SAP Innovation Center

Jan 2017 - Dec 2018 (1 year 11 months)

Extended, profiled, tested, and optimized a computer-vision inference engine with custom CUDA kernels, processing Full HD video at up to 500 fps on a DGX-1. Rewrote an algorithmic bottleneck in OpenPose, improving full-model performance from 10x to 70x over the original.

CUDA Computer Vision Inference OpenPose Profiling Optimization Neural Networks