Bedilbek Khamidov

Co-founded a cashierless retail startup, leading four engineering workstreams and up to 15 engineers; personally hired ~90% of the team.

Built a real-time C++ multi-camera tracker that fused Kafka 2D poses via epipolar geometry, clustering, triangulation, and joint optimization into tracked 17-joint 3D skeletons. Optimized ONNX/TensorRT and FP16/INT8 inference to scale from 10 to 40 camera streams at 30 FPS per RTX 6000 Ada, with up to 5% relative AP loss.

Led synchronized video capture and calibration for 150 cameras. Developed a CUDA/C++ bundle-adjustment solver, cutting extrinsic calibration from days to hours; introduced automation reducing the wider process to minutes.

Built cloud/on-prem GPU infrastructure, monitoring, and a cost-aware pipeline handling up to ~3 TB/day. Developed ML data, annotation, synthetic-data, and Unity digital-twin workflows, supporting action recognition with 21K public and 50K proprietary clips and tools for debugging shopper interactions.

Designed and implemented the initial Python/Django backend and pharmacy back office, establishing multi-tenant architecture and cloud deployment foundations before handing development over to other engineers.
Moved complex reporting aggregates from application-side processing into PostgreSQL, reducing database round-trips and improving report responsiveness.

Co-founded a software development company and led delivery of three products with a team of 10 engineers; assigned and reviewed engineering tasks and mentored team members in GitLab CI/CD and project management.
Texnorama
Uzbek-language technology news platform. Led DevOps.
Partiya Edi
Food delivery product. Led backend engineering and DevOps.
Avtovin
Automotive classifieds product. Led backend engineering and DevOps.

Defined architecture, Agile workflows, infrastructure, and delivery standards; established GitLab CI/CD to reduce release overhead.
Implemented Docker/Compose deployment, Gitlab CI/CD and registry push/pulls across both projects
Trucking Management System
Architected and deployed a multi-tenant SaaS Trucking Management System to automate dispatcher and driver workflows.
Amazon Relay Automation
Built a Go/Kafka service to automate Amazon Relay load matching and booking, using concurrent workers, configurable filters, and a booking queue prioritized by price per mile.
Configured multiple IP addresses on an AWS EC2 instance and implemented per-IP HTTP connection pools with source-address binding, response-driven throttling, and request timeouts.
Added Prometheus metrics for request latency, booking delay, filtering throughput, order status, and request counts by company/IP for Grafana monitoring.
Implemented Datadog monitoring for containers, processes, and logs.