Building production AI systems — not just slides about them.
Staff-level AI/ML Engineer with 10+ years shipping code in production. I built my expertise the hard way — training models, debugging distributed pipelines, and pushing inference to the edge. In the USA tech startup ecosystem, I operated as a Staff Engineer: the person who codes the hardest parts, mentors the team, and still reviews CUDA kernels at 2 AM.
My track record spans from zero-to-one AI products (text-to-video GenAI, autonomous LLM agents) to scaling systems under real user load. I've worn the founder hat, but my core identity remains technical — I architect systems I can also implement end-to-end.
Model Development & Training
- Fine-tuning LLMs (LoRA, QLoRA, RLHF) and deploying them for inference optimization (vLLM, TensorRT-LLM)
- Computer Vision: synthetic data generation, GANs/Diffusion models, face/body keypoint estimation
- Classical ML at scale: XGBoost, feature stores, real-time inference pipelines
MLOps & Infrastructure
- End-to-end ML pipelines: data versioning (DVC), experiment tracking (MLflow, W&B), CI/CD for models
- Distributed training on multi-GPU clusters (PyTorch DDP, DeepSpeed)
- Production deployment: Docker, Kubernetes, model serving with FastAPI/gRPC, edge optimization (ONNX, TensorRT)
Full-Stack System Architecture
- Building AI-native applications: vector DBs (Pinecone, Milvus, pgvector), retrieval-augmented generation (RAG)
- Cloud-native stacks: AWS/GCP, serverless GPU inference, event-driven architectures
- Low-level optimization: CUDA profiling, memory-efficient attention, custom PyTorch operators
Core Competencies: LLMs & GenAI (training + inference), Computer Vision, MLOps, Distributed Systems, Technical Strategy, Team Mentorship
Languages: Python (expert), C/C++ (systems/performance), SQL (data at scale), Russian (Native), English (Professional)
|
|
|
|
- 🏅 Starta Accelerator Alumni (New York City, 2020) – Capital raise for NeuraLoom.
- 🎯 iDealMachine Award / GoTech (2016) – U.S. acceleration grant for VR/AR training platform.
- 💡 Top 50 Innovation Ideas (2014) – HEATCAM (Real-time infrared diagnostics).
- 🏆 Intel / APPS4ALL Hackathon Winner (2013) – Best Android App for Intel x86 architecture.
- 🎓 B.S. Information Systems and Technologies | Ural Federal University (2007 - 2011)
- ⚡ Engineer's Degree in Electrical Science | Polytechnic College (2000 - 2004)
"Good engineers write code. Great engineers train models that write code."