CV

ALIKHAN KABDUALY

Building AI products at 16. From Kostanay, Kazakhstan.

The same content as the PDF. This page prints onto a white sheet — no glass, no light cone.

About

I'm Alikhan, a 16-year-old developer from Kostanay, Kazakhstan. I build AI-powered products — computer vision models, progressive web apps and Telegram bots — taking them from concept to deployment. I've competed in hackathons and pitch sessions across the Astana Hub ecosystem in Kostanay, Petropavlovsk and Almaty. In September 2026 our team took first place at the Qostanai AgroTech Hackathon with AgroVision AI, a weed map built from drone imagery. Right now I’m building a new project — no details yet.

Projects and venues

  • AgroVision AI2026-09 · 2 days + post-event work

    A weed map and spot-spraying plan from drone imagery

    Venue: Qostanai AgroTech Hackathon 2026 · case No. 1 · 1st place

    Role: Development and ML — the detector and classifier, the agronomy rules, georeferencing and task export for machinery, the server, the agronomist's offline app and the dashboard

    Stack: Python · PyTorch · YOLOv8s · EfficientNet-B0 · FastAPI · Streamlit · PWA · GeoJSON · ISO 11783-10 · Docker

    • 0,77 detector mAP@50 on an independent test, 381 images
    • 26 weed species and 3 stages in the classifier
    • 98,4% species accuracy on 1,591 reference photos
    • 500 000 ₸ prize for first place
  • Jasyl2026-08 · 2 days

    AI monitoring of Kostanay's urban trees

    Venue: Qostanai Smart City Hackathon · case No. 1

    Role: Development and ML — the app and its cross-platform builds, model training, the inference service, the database and its access rules, offline sync

    Stack: Vite · React 18 · TypeScript · PyTorch · MobileNetV3-Large · ONNX Runtime Web · Supabase · Leaflet · Capacitor 6 · Electron 33

    • 343 photos in the dataset, split by tree 268 / 75
    • 54,7% best validation accuracy
    • 0,55 macro F1 across six classes
    • 16,8 МБ ONNX model for CPU inference
  • QA Vision2025-11 · 3 days

    Paint-quality control on car bodies, from a photograph

    Venue: Qostanai AI-Sana Industry Hackathon · Allur Challenge

    Role: Development and architecture — the dashboard, the login service and API, the pluggable AI module system, the detection registry and KPI maths, the exports, roles and access control

    Stack: Python · Streamlit · Plotly · FastAPI · Roboflow · Ultralytics YOLO · PyTorch · OpenCV · pandas · PyJWT · Power BI · Looker Studio

    • 7 787 lines of Python in the project
    • 4 kinds of AI module behind one interface
    • 0,70 default confidence threshold
    • 3 access roles: superadmin, plant manager, worker
  • KZ UniVerse2025-12 · 24 hours

    A single platform for Kazakhstan's universities

    Venue: IT Fest 2025 · Hackathon track

    Role: Development — the catalogue and university pages, comparison, analytics, the AI provider chain, profile and portfolio, the Telegram bot, and the demo rig for the pitch

    Stack: Next.js 14 · TypeScript · Tailwind CSS · Framer Motion · Recharts · Ollama · Google Gemini 2.5 Flash Lite · Telegram Bot API

    • 15 universities in the base
    • 19 study programmes
    • 3 assistant modes with fallbacks
    • 24 ч for the whole track
  • NPAI2025-12 · Since 2024, across several venues

    A marketplace of ready-made industrial IoT modules

    Venue: Startup Battle, IT Fest 2025 · top 30

    Role: Product and development — the marketplace front, the dashboards, the Eco module, login and protected code preview, the project economics, the pitch and the answers to the jury

    Stack: JavaScript · HTML/CSS · Google OAuth · IoT · Power BI

    • 30 teams in the in-person Startup Battle final
    • 3 cities where the project was shown
    • 2024 the year of the first venue
    • 4 revenue streams in the model

Timeline

  • Idea battles and pitch sessions at Qostanai Hub2024 · Kostanay

    First outings with NPAI: learned to fit an idea into a short pitch, and to hear questions about money rather than code.

  • Pitch venue2024–2025 · Petropavlovsk

    A conversation with people from outside my own city, where the question of who actually buys this lands hardest.

  • Qostanai AI-Sana Industry Hackathon · Allur Challenge2025 · Kostanay

    1–3 November at Dulatov University. The first product built against a real manufacturing case — QA Vision.

  • IT Fest 2025 — two tracks2025 · Almaty

    5–6 December. A 24-hour hackathon with KZ UniVerse and Startup Battle with NPAI: top-30 teams, pitched in person.

  • Qostanai Smart City Hackathon · case No. 12026 · Kostanay

    13–14 August. Jasyl: our own dataset, a fine-tuned model and a working offline app in two days.

  • Qostanai AgroTech Hackathon · 1st place2026 · Kostanay

    17–18 September. AgroVision AI: a weed map from drone imagery for Olzha Agro — first place on the case and a 500,000 ₸ prize.