Open to opportunities

Антон Агейков

@0005282

Message

I build high-performance computer vision and LLM inference systems for resource-constrained hardware.

Russian Federation

Message

What I'm looking for

I'm looking for remote or relocation opportunities where I can build high-performance AI inference and computer vision systems, optimize models for constrained hardware, and keep learning deeply.

At Huawei, I develop CPU inference systems for LLMs, including SIMD-vectorized fused C++ transformer attention operators that improved end-to-end throughput by 20%. I've also ported vLLM fused operators while preserving a 5x throughput gain and built vLLM, llama.cpp, Lmcache, and Valkey server infrastructure for AI agent coding workflows.

Previously, I trained and deployed YOLO, UNet, and vision transformer models for real-time vision pipelines at Luggar Design Bureau and Neurolumber. My work spans RKNN quantization, object tracking, shared-memory multiprocessing, service-based architecture, ZeroMQ integration, and GPU/NPU optimization, with outcomes including reducing YOLO inference from 300 ms to 40 ms and achieving 95% mAP.

Experience

Work history, roles, and key accomplishments

Current

CPU Inference Specialist

Current

Huawei

Feb 2026 - Present (6 months)

Developed fused C++ operators for transformer attention via SIMD on ARM, achieving 20% throughput improvement. Ported operators to vLLM, preserving 5x throughput gain, and set up vLLM and llama.cpp servers for AI agent coding.