AI PRODUCT ENGINEER
I turn product goals into working systems — from architecture through validation and delivery.
Python backends · AI-assisted systems · Data pipelines · Telegram products · Web interfaces
Open to remote roles with stable monthly employment and daily written communication.
A Telegram Mini App and Python backend for marketplace deal intelligence on Kufar.
- Problem: Fresh listings need to be normalized, compared with market baselines, analyzed for risk, matched to purchasing profiles, and delivered through a consistent product state.
- Ownership: Product definition, market research, architecture, Python backend, data pipeline, AI-assisted analysis, authenticated API, Telegram Mini App, deployment, validation, and release boundaries.
- Evidence: The production MVP reached an owner-provided pipeline snapshot dated 2026-07-17: 2,993 stored listings, 2,494 evaluations, 3,258 successful AI analyses, and 555 Telegram notifications. These figures describe pipeline volume and engineering maturity, not traction or revenue.
- Links: Case study · Live product · Telegram bot
A local tool for reasoning about changing university admission tables without presenting uncertain source data as exact.
- Problem: University sources use different table layouts and scoring conventions, while source schemas can change without notice.
- Ownership: Parser pipeline, schema validation, priority-cascade model, uncertainty policies, Quickshell interface, observable source states, and rollback-aware installation.
- Evidence: Public, installable source with fail-closed schema checks, visible uncertainty states, demo screenshots, configuration guidance, and a reversible update path.
- Links: Case study · Source and documentation
A privacy-conscious Quickshell module that turns local Codex session data and rate-limit snapshots into a compact usage panel.
- Problem: Usage data arrives from different local sources and at different freshness levels; a number without source state can appear more authoritative than it is.
- Ownership: Data collector, cache and fallback states, privacy controls, QML interface, theme behavior, settings, and installation path.
- Evidence: Public, installable source with live, cached, stale, and local-only states; workspace-name collection is opt-in, and collection pauses when the module is disabled.
- Links: Case study · Source and installation guide
I work best with a clear outcome, concrete constraints, and responsibility for a complete system or finished product slice. My usual scope connects product logic, architecture, Python and data workflows, interfaces, validation, and delivery.
Building Hmara and F-Signal. This profile, portfolio, and hiring contact represent my individual work and responsibility.
AI Product Engineer
Превращаю продуктовые цели в работающие системы — от архитектуры до проверки и доставки.
Backend на Python · Системы с ИИ · Конвейеры данных · Telegram-продукты · Веб-интерфейсы
Рассматриваю удалённые роли со стабильной ежемесячной оплатой и ежедневной письменной коммуникацией.