Lead CRM and retention
We do not publish candidate contacts. Email, Telegram, LinkedIn — in the original CV on @cyprusithr (Telegram MTProto) ↗
- Role
- CRM and retention
- Grade
- Lead
- Traffic
- native
- Reach
- direct
- Platform
- @cyprusithr (Telegram MTProto) ↗
- Updated there
- 2026-07-29 stale
- In our base since
- 2026-09-13
As it arrived from @cyprusithr (Telegram MTProto) · fetched 2026-09-13 — the candidate's own text, contacts removed.
#CV #resume #ml #ai #computer_vision
#CV #resume #ml #ai #computer_vision
Hi everyone!
I’m Georgy Gunkin - Lead ML/CV Engineer **(5+ years)** building production, real-time computer vision systems end-to-end: problem framing, data strategy, training, GPU inference optimization, deployment, and production monitoring. Performance-first mindset (NVIDIA Nsight Systems/Compute), reliability-first operations (metrics, monitoring, reproducibility).
Core strengths
• End-to-end ownership of ML/CV systems in production (from data + training to rollout + monitoring)
• Real-time performance engineering: multithreaded/asynchronous pipelines, profiling, GPU-accelerated inference (**TensorRT **/** ONNX Runtime **/ **CUDA**)
• Production engineering: **Rust** / **C++** / **Python** / **C#**, **Docker/Linux**; microservices and monoliths
• Leadership: led ML subsystem development and managed a team of 3 ML engineers
Selected production highlights
• Diffusion-based clothing try-on: optimized pipeline for speed/quality, achieving **420%** speedup over baseline; built end-to-end LoRA training and model distillation pipelines
• Lipstick virtual try-on: designed and implemented a service handling **1,000+ RPM**, now serving 100% of sellers on the platform
• Produce inspection conveyor: **2** industrial cameras (**4096x3000** @ **23.5 FPS**), detection/segmentation + defect pipeline, track-level decisions, stable at full conveyor load
• Retail theft-risk analytics: distributed services (detection/segmentation/pose + multi-camera tracking/ReID + action recognition), **40** **1920x1080** streams at **6 FPS** with low latency
• NDT for metal products: **4** cameras (**2448x2048** @ **79 FPS**), **14** product types / **9** defect classes, active-learning data loop + operator GUI
I’m currently exploring Senior/Lead opportunities where I can own the end-to-end lifecycle of ML/CV systems - from problem framing and data strategy to GPU-optimized deployment and production monitoring.
Open to Senior/Lead ML/CV roles (CV Tech Lead / real-time video analytics / inference & performance engineering).
Languages: English (fluent), Russian (native)
**Full CV: **[contact hidden]
**Contacts**
[contact hidden]
[contact hidden]
[contact hidden]