About
I’m Baktiyar. I build things that learn, then find out where they stop working.
I’m from Kazakhstan, in grade 10 at the Nazarbayev Intellectual School of Physics and Mathematics in Astana. I work across software, small hardware and research: machine-learning systems, microcontroller devices, and studies that end in a manuscript or a negative result.
Much of it started at a competition or a hackathon; the rest are questions I followed on my own. In grades 7 and 8 I was on the school physics team.
What I work on
01Machine learning
Most of my work. Lately less about bigger models than about evaluation: data from a new source, a model under several turns of pressure, knowing when to abstain.
02Small hardware
A Transformer designed to fit a microcontroller’s memory; a classifier compiled into a wearable’s firmware.
03Security
I practise CTF challenges on picoCTF and Hack The Box, work through OverTheWire Bandit, and study Linux, web security and security fundamentals.
04Physics & the sky
My longest-running interest: a club I co-founded and an international competition.
How I try to work
Measured, derived, planned
Keep them apart. The NanoTranslate preprint leaves board latency and BLEU blank rather than estimate them.
See it →Keep the negative result
SunCast’s audit did not confirm the gain it was looking for, and the record says so.
See it →Exact where it matters
The Halyk engine computes covenants with decimal arithmetic and records where every number came from.
See it →
Outside the record
- Mostly MatterA short-video channel where I explain physics and engineering. More than 4,200 people follow it.
- ChessI play at first category, and I’ve spent close to two hundred hours teaching younger students.
- QuasarThe school astronomy club I co-founded: lectures, research sessions, and guests from Nazarbayev University.
- VolunteeringEvent shifts in Astana — rocket-launch preparation with KazRockets, matches at Barys Arena.
Currently learning
- PCB design
- Embedded systems, past the dev board
- Computational physics
Reading & self-study
- Mathematics
- Stewart’s Precalculus and Calculus, linear algebra, complex numbers, an online course from HKUST.
- Physics
- Mechanics, astrophysics, and computational physics.
- Courses
- Astronomy: Exploring Time and Space, and Introduction to Psychology (Coursera, 2025).
- A book I return to
- The Information, by James Gleick.
Education & languages
Nazarbayev Intellectual School of Physics and Mathematics
Astana-Nura · Grade 10 (2026–27) · expected graduation 2029
Kazakh (native) · Russian (native) · English (proficient)