Architect & Research Scientist · Real-Time AI Systems
Armen is an architect and research scientist with over a decade of experience turning cutting-edge AI research into scalable, real-time systems. He builds at the intersection of performance-critical infrastructure and AI — from cloud video analytics processing 20M+ camera streams, to autonomous driving stacks for 100+ vehicles, to LLM-powered products serving real users.
As Tech Lead at Sber Automotive Technologies he supervised the prediction and perception teams behind 100+ self-driving cars and trucks, extending object detection range from 50 m to 200 m and cutting 3D tracking latency from 100 ms to 5 ms. At Network Optix he drives architecture across real-time video analytics, AI inference, vector search, agentic LLM tooling and cloud storage. His roots are in robotics research: drones, wall-climbing and capsule robots — with a PhD in Physics & Mathematics from MIPT and 10+ publications in WoS/Scopus.
From Research to Real-Time Production
Armen specialises in translating ambiguous problems into shipped products — designing end-to-end systems from data pipelines and ML inference to cloud infrastructure and cost optimisation. His track record includes 20× performance gains, 50%+ cost reductions and a 99.5% reduction in RPS load.
He built a proprietary Claude Code development harness with plugins, prompt evals and QA agents, and pairs hands-on depth across C++, Python, Go, AWS/GCP/Azure and ML/AI stacks with experience growing and mentoring engineering teams from 3 to 15 engineers.