1st place
AEROO Space AI Competition
First place with TRACE, a maritime monitoring system built on satellite and AIS data, at the international competition “Exploring the World of Science”.
Baktiyar Ablaikhan
I build small machines that learn — a translator sized for a microcontroller, a wearable fall detector, satellite monitoring for the sea — and test how far they can be trusted.
Grade 10, NIS Physics and Mathematics, Astana. School physics team in grades 7–8. Now: machine learning, embedded systems, and security through CTF practice.
Every dark line is a piece of work.
Work
All work →- TRACE — Satellite imagery and ship transponder (AIS) data in one tool for vessel detection, oil-spill segmentation and anomaly flags.1st place, AEROO Space AI Competition
- SunCast — Machine-learning solar-flare forecasting from NASA and NOAA satellite data and multimodal space-weather signals.97.07% accuracy, 95.26% recall · with Dr. Dana Alina, Nazarbayev University
- EpiCare — A wearable on the ESP32-S3 that detects falls on the device itself, without sending raw data to a server.98.9% of falls caught, on the device
- Mitra — Digital SAT practice that explains every mistake, a vocabulary trainer, and a catalogue of opportunities for students.
Honours
All honours →Gold Honour
International Astronomy & Astrophysics Competition
19 points in the supervised 20-question final; top 2% of all participants.
Grand finalist · Top 25
NSRI Summer Research Hackathon
Grand finalist with GraphTrust, an explainable graph analysis of multi-hop identity risk.
75th of 4,370
Kaggle: Predicting Heart Disease
A tabular binary-classification competition, ranked on the final private leaderboard.
Ahead of 98.3% of 4,370 teams
Research
All research →- NanoTranslate-S3 — Offline Russian-to-English neural machine translation designed around an ESP32-S3 microcontroller.
- Sycophancy as a pathogen — A pilot study of how language models drift toward agreeing with a user under sustained pressure, and whether DPO reduces it.
- Corrosion prediction across publications — Predicting the pitting potential of multi-principal-element alloys from literature data, and testing whether the model knows when not to trust itself on a new publication.
- GraphTrust — Typed attack-path reasoning for cloud identity permissions, with remediation that keeps business workflows intact.
- ArtEco — A small artificial economy where agents learn to gather, barter and signal without prices or a language given to them.
- Deep-SAR Oil Spill Segmentation (Refined) — Satellite radar images of the sea with oil-spill masks corrected by hand — 38% of training masks and half of the validation masks.Cited: used in a peer-reviewed paper in Discover Computing (Springer Nature, 2026)
Kazakh Classical Poetry — Abai & Makataev — 233 cleaned Kazakh poems by Abai Kunanbaiuly and Mukagali Makataev, prepared for language models.
Community
All community →
Mostly Matter
1.1M+ views across 26 videos
Astana, now
Quasar Astronomy & Astrophysics Club
Co-founder
Chess
190+ hours teaching younger students
Volunteering
Event volunteering in Astana, including NIS Run Fest 2024, KazRockets and matches at Barys Arena.
Experience
