Skip to content

17-year-old Undergraduate Student based in Astana, Kazakhstan

Amir Nagiyev — AI & ML Engineer & Systems Architect

Building neural networks on top of a low-level mindset — from tensor graphs in PyTorch to zero-allocation hot paths in Go.

Explore Projects
Focus
AI / ML
Core lang
Go · Py
Target
JLPT N2

Spotify: IJWW

01 / About

Learning machines, understanding metal.

I approach AI the way a systems engineer approaches a runtime: measure, profile, and understand every layer before trusting it.

Academic focus

Undergraduate in Artificial Intelligence & Machine Learning. Studying how models learn — optimization, representation and the architectures behind modern neural networks.

  • Deep Learning
  • Optimization
  • Linear Algebra

Linguistic profile

Fluent in Russian and English. Actively learning Japanese with Genki → Renshuu stack, targeting JLPT N3 then N2.

  • RU · Fluent
  • EN · Fluent
  • JA · Learning

Systems mindset

Obsessed with what happens beneath the abstraction: Go memory mechanics, stack vs. heap allocation, struct field alignment and padding — and how the same rigor applies to network design.

  • Stack vs Heap
  • Struct Alignment
  • Escape Analysis

02 / Skills & Architecture

The stack, from tensors to syscalls.

Tile 01

AI & Machine Learning

Exploring neural networks and data pipelines through undergraduate study and hands-on experiments with Python and PyTorch — from data preparation to model training and evaluation.

example · epoch 0loss ↓ 0.0412epoch 64
  • Python
  • PyTorch
  • Data Modeling
  • Model Training
  • Model Evaluation

Tile 02

Systems & Backend

$ go build -gcflags="-m" ./...
./node.go:14: &Node{} does not escape
./pool.go:31: 0 allocs/op
  • Go
  • Pointer Rewiring
  • Escape Analysis
  • Zero-Alloc
  • FastAPI
  • Python
  • SQLite WAL

Tile 03

Web Ecosystem

  • Next.js 15
  • TypeScript
  • Tailwind CSS
  • Framer Motion
  • WebGL
  • Spline

Tile 04

Languages & Tools

  • VS Code
  • Git

RU ●●● EN ●●● JA ●●●

03 / Projects

Voyage AI MVP

An asynchronous travel recommendation engine for discovering places across Kazakhstan.

MVP · Livev0.3.1

Voyage AI

A fully async FastAPI service that ranks destinations by user preferences. Strict Pydantic v2 models validate every request, while SQLite in WAL mode allows concurrent reads during writes — keeping tail latency flat under load.

  • FastAPI
  • Pydantic v2
  • SQLite WAL
  • async / await
  • 60+ seeded entities
p99 latency
< 45ms
Seeded entities
60+
Regions
KZ-wide
Layout
Responsive
GitHub RepositoryLive Demo
GET /v1/recommend?vibe=nature200 · 38ms
  1. 01

    Charyn Canyon

    Almaty Region

    0.97
  2. 02

    Burabay National Park

    Akmola Region

    0.94
  3. 03

    Baiterek Tower

    Astana

    0.91

latency.p50 12ms

latency.p95 31ms

latency.p99 44ms

04 / Contact

Open a session.

Prefer the command line? Query the node directly — or type socials to reach me on Telegram, GitHub, LinkedIn or Freelance.kz.

amir@astana-node:~

astana-node v1.0 — connection established.

Type help or click a command below to begin.

amir@astana-node:~$