Open to speaking
// Senior Software Engineer · High-Performance Systems
I build systems where every millisecond and every byte counts: video cores, face recognition engines, computer-vision pipelines and production AI agents.
TOO “CIT” Orleu, Karaganda
open source · Apache 2.0 · WebAssembly
A full face stack that runs entirely in the browser — detection, 98-point landmarks, 576-point 3D mesh, recognition and passive anti-spoof. Models trained from scratch; zero server, camera frames never leave the device.
LFW 99.07%3.0 ms / faceC · WASMAES-256-GCM weights
computer vision · ANPR / LPR · 2025
Own neural pipeline for automatic license-plate recognition of Kazakhstan plates (and 6 more countries) plus vehicle-type classification — for parking, barriers and access control. Detect → rectify → read → classify, served over a REST API. Co-authored with Aidos Zhumadilov.
40–60 ms / frame (CUDA)7 countriesCRNN + CTCPyTorch · OpenCV · ONNX RuntimeGo · C helpers
open source · MIT
Zero-build reactive UI engine for server-rendered pages: signals, islands, html`` templates, forms, router — one ES module, 0 dependencies. This site runs on its signals core.
video
Video management stack built for high-efficiency bitstream processing (H.264 / H.265) without heavy pixel-decoding overhead.
now
Backend logic for autonomous systems, automated analytical loops and smart video-processing agents.
| Aegis | Alpine.js | htmx | Petite-Vue | Stimulus | Preact islands | |
|---|---|---|---|---|---|---|
| Build step | none | none | none | none | none or bundler | usually (none via htm) |
| Size, gzip | 7 KB signals · 30 KB islands+templates | ~15 KB | ~14 KB | ~6 KB | ~11 KB | ~4 KB core + addons |
| Reactivity | signals, computeds, effects | proxy + directives | — (server round-trip) | proxy + directives | — (imperative) | hooks / signals add-on |
| Templates | html`` in JS, CSP-safe, keyed list() | x-for / x-text in HTML | HTML from server | v-for / {{ }} | targets in HTML | JSX / htm |
| Data cache | resource(): ETag, revalidate, offline | — | per-attribute requests | — | — | — (3rd-party) |
| Forms | validation, schemas, 422 mapping | — | server-side | — | — | — (3rd-party) |
| Router | Navigation API, boost(), View Transitions | — | hx-boost | — | via Turbo | separate package |
Aegis figures from its README (1085 tests, 0 dependencies, works with Django / Laravel / Rails / Go / PHP). Others from their public docs; sizes approximate. Alpine and Petite-Vue are smaller and great for sprinkles; htmx is ideal when the server should do everything — Aegis targets pages that need real client state, data and forms without a SPA.
Open to conference talks, workshops, meetups and podcasts. Topics come straight from systems I’ve built and run in production.
Go · C (C99) · Python · JavaScript / Node.js · PHP · SQL
Face recognition & biometrics · NN inference optimization · Edge AI · RAG · LLM agents
RTSP / ONVIF · H.264 / H.265 bitstream transformation · Streaming & video cores
PostgreSQL · MySQL · Firebird · SQLite · MongoDB · Docker · Linux
Vue.js · React · Express · Electron · Sequelize · Yii / Yii2 · Aegis
Kazakh — native · Russian — native (C2) · English — B1, technical
Education — B.Sc. in Engineering (CAD), Karaganda State Technical University, 2006
🎙 Available for talks and conference speaking on AI & computer vision — invite me.
WhatsApp +7 708 356 7955 Call Telegram @atynov LinkedIn GitHub CV (PDF)