> For the complete documentation index, see [llms.txt](https://www.saalse.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.saalse.com/computer-vision-ai.md). # Computer Vision AI For over a decade, I’ve been taking computer vision products from 0 to 1. From sub-second real-time streaming to heavy industry safety, manufacturing logistics, and spatial AI, my focus is on solving complex physical-world problems through edge optimization, precise pipeline architecture, and building world-class R\&D teams. Here is a look at my hands-on experience and architectural work across various domains: ### Spatial AI & 3D Reconstruction **2026** I provide business and solution consulting, as well as R\&D team building, for Fanis, an elite AI lab founded by Nick Falaleev, one of the world's top computer vision engineers, a Kaggle Master and CVPR author. Together with ex-Apple AR/VR experts, we're building ambitious tech that turns any smartphone into a precise 3D scanner, even without LiDAR. ### Sub-Second & Real-Time Video Analytics in Sports Tech **Online Game Tracking (2025)** As a Solution Architect and R\&D Team Lead, I went hands-on with the coding and ML to build a sub-second, real-time vision AI pipeline on cloud GPUs. I engineered a FastAPI Websocket image receiver that recognizes rapid gaming screenshots using a fusion of OpenCV, YOLO (custom-trained for 53 classes), and PaddleOCR optimized on TensorRT. This pipeline outputs to a signaling API with JSON game tracking data, achieving 99.9% accuracy. **OSAI (2019-2021)** I provided product design consulting and helped build the R\&D team for a high-speed sports analytics platform capable of processing live tracking at 120 FPS for professional coaching and sport markets.
### Real-time Video Analytics for Legacy Industries **Heavy Equipment Manufacturing (2025, Heavy Industry)** As a Solution Architect, I consulted a heavy equipment manufactory on an edge-computing CV system to monitor worker productivity and ensure safety across heavy machinery plants. The pipeline tracks tool usage, operation cycle timing, PPE compliance, and dangerous zone intrusion detection to drastically reduce downtime and safety violations. **Interquell (2025, Food Production)** I designed a comprehensive Vision AI pilot concept to automate packaging and internal logistics for dry pet food production. The pipeline visually tracks the product journey from "bag to truck", featuring real-time seal quality inspection, 360-degree pallet scanning, forklift tracking, live inventory mapping, and dispatch mismatch alerts to prevent loading dock errors.
**Lima Labs (2025, Agriculture)** I worked as a Product Design Consultant to help shape Lima Labs' computer vision products, which focus on highly accurate product yield forecasting and analytics for commercial farms in Kenya.
**Integra (2020-2021, Oil & Gas)** As Solution Architect and R\&D Team Lead, I designed and built a full pipeline PoC aimed at preventing emergencies and decreasing human-factor risks on oil and gas drilling rigs. The system captured live CCTV RTMP streams and processed them through a deep learning CNN model to recognize specific equipment states and accurately count rig pipe screwing turns, feeding everything directly into a real-time incident dashboard.
**Waste Management Consulting (2021, Urban)** As a Consultant, I designed a computer vision PoC capable of accurately identifying trash bin fill levels. The solution was built to optimize collection routes and improve operational efficiency for urban waste management operations. **AiTend (2020, AgTech)** As a Solution Consultant, I helped design a computer vision system for dairy farms that utilized advanced pose estimation to detect unique cows and continuously monitor their health states and rumination periods.
### Satellite Imagery Analytics **Agro Hack (2020)** As Team Lead during a 40-hour competition, I guided a team to develop a machine learning solution utilizing satellite imagery to determine agricultural lands prone to waterlogging. We built a production-ready web application that allowed users to select an examination area and export visual computation results as GeoJSON. The computational algorithm utilized segmentation analysis on top of a U-Net architecture for Sentinel satellite imagery tiles (using NIR, SWIR1, and RED bands), returning the results as a binary mask.
*** ### **Facial Recognition** **Facepunk (2022–2023, Entertainment & Healthcare)** A Computer Vision Advisor for AI startup focused on deep facial data, I provided ML/Deep Learning expertise and initiated product development. *** ### My Very First CV Project **Nestle & Danone (2013, Advertisement)** My practical work with computer vision dates back to 2013. As a Solution Architect and R\&D team lead, I designed and built interactive gaming stands deployed in shopping malls for Nestlé and Danone. We used a combination of Xbox Kinect cameras, custom drivers, and OpenCV pose estimation to create interactive games that engaged thousands of users with these brands.