AI Insights

Technical deep-dives, industry analysis, and product updates from the Ebsar team.

Computer Vision

Building a Shelf Recognition System: From YOLO to Production

A technical walkthrough of how we built Ebsar Merch's core recognition engine — from dataset creation and model training to deployment at scale across 200+ stores in Kazakhstan.

Aigerim Muratova

15 Mar 2026 · 12 min read

Computer Vision

YOLOv8 vs Detectron2: Benchmarks on Kazakhstani Retail Data

We tested both architectures on our proprietary dataset of 50,000+ shelf images. Here's what we found about speed, accuracy, and deployment trade-offs.

Aigerim Muratova

8 Mar 2026 · 8 min read

Product Updates

Ebsar Merch v2: What's New

Price tag recognition, multi-brand planograms, and a completely redesigned dashboard. A walkthrough of everything in the latest release.

Madina Ibragimova

1 Mar 2026 · 5 min read

Industry

The State of FMCG Distribution in Central Asia (2026)

An analysis of distribution challenges, technology adoption, and the growing role of AI in merchandising across Kazakhstan, Uzbekistan, and Kyrgyzstan.

Amir Suleimanov

22 Feb 2026 · 10 min read

Engineering

How We Built a CV Pipeline Inside a Telegram Bot

Our agents use Telegram to submit shelf photos. Here's how we built the real-time image processing pipeline that powers it.

Daulet Kenzhebaev

15 Feb 2026 · 7 min read

Open Source

Open Sourcing Our Kazakhstani License Plate Model

We're releasing our KZ plate detection and OCR model under MIT license. Trained on 20,000+ images of Kazakhstani plates in real-world conditions.

Bakyt Nurlanuly

8 Feb 2026 · 6 min read

Engineering

Deploying CV Models on NVIDIA Jetson for Retail Cameras

Running inference at the edge: how we optimized our retail analytics models for NVIDIA Jetson devices to achieve real-time processing.

Nursultan Akhmet

1 Feb 2026 · 9 min read