Product builder / Delivery Driver

Build.
Automate.
Simplify.

I turn vague problems and awkward processes into small working products — with AI, automation and code.

Product → prototype AI-native workflow

Capability field

Portfolio Query local composer

portfolio

Ask me about Timur, his work, projects, product approach or technical decisions.

Runs entirely in your browser. It uses a small local knowledge base, fuzzy matching and intent scoring — not an LLM. Press / to focus.

Selected work

A consulting and process-redesign project for a manual communication-quality audit workflow.

What: Mapped the workflow and moved manual Google Workspace steps into an internal CRM.
Why: Reduce fragmented tracking and create a cleaner base for future automation.
Detail: Researched external audit-automation solutions and evaluated options for further development of the process.

Development of a B2B loyalty CRM product spanning web, mobile, backend and external integrations.

What: Requirements, backlog, planning and delivery across a multi-application CRM ecosystem.
Why: Give business customers one system for managing loyalty programs and connected workflows.
Detail: SaaS and on-premise product with four web apps, two mobile apps, PostgreSQL and integrations.

Operational dashboards that turned fragmented process data into a shared view of performance and bottlenecks.

What: Built SQL + BI reporting for operational and product metrics across 20 products.
Why: Replace repetitive manual reporting and make deviations visible earlier.
Detail: Reduced the time spent preparing regular reporting by about 70%.

An n8n/JavaScript workflow that automates invoice preparation for remote employees.

What: Validated source data and generated ready-to-use invoice documents from a repeatable workflow.
Why: Remove repetitive manual preparation and reduce the risk of document errors.
Detail: One invoice was generated in under 15 seconds and the workflow was used in real monthly cycles.

An AI-assisted book-translation system with durable progress and a Telegram interface.

What: Stores translation progress, terminology, style and review state outside a single chat session.
Why: Long-form translation needs consistency and the ability to continue across multiple sessions.
Detail: The Telegram bot is the user-facing layer of the same translation workflow.

An automated pipeline for producing short-form video from research and scripting through rendering and review.

What: Connected n8n, LLM steps, TTS, subtitles and FFmpeg rendering into one workflow.
Why: Reduce repetitive handoffs and make content production more repeatable.
Detail: Generation is separated from publishing so the final output remains reviewable before release.

A Chrome extension that shows movie and series ratings directly in the YouTube viewing context.

What: Identifies the title, matches it deterministically and renders the rating next to video metadata.
Why: Remove the extra search step when deciding whether a film or series is worth watching.
Detail: Runs without its own backend, LLM or required API key; browser behavior is covered by automated smoke tests.

A browser product for automating repetitive parts of the job-search workflow on hh.ru.

What: Combines AI-assisted vacancy handling, cover letters, automated responses and resume-performance analytics.
Why: Reduce manual job-search work while keeping the process measurable.
Detail: The product also has its own backend layer for tariffs, licensing and payments.

A data product for collecting and comparing labor-market signals across professions.

What: Combines a browser collector, validated snapshots, history and a web interface for analysis.
Why: Make profession and competition research reproducible instead of relying on one-off manual checks.
Detail: The system flags incomplete or invalid observations and keeps the actual collection date so market results can be compared correctly over time.

A lightweight ingestion utility that turns RSS/Atom sources into predictable content for AI agents.

What: Normalizes inconsistent feed formats into a stable structure for downstream processing.
Why: Agents should not need to reimplement feed parsing every time they consume a new source.
Detail: Designed to stay small, dependency-light and portable across Windows, macOS and Linux.

About

I like software that removes friction instead of adding another system people have to manage.

My background is in project and process management, so I usually start from the workflow, constraint or user problem — not from the technology.

Now I work hands-on: prototypes, AI agents, APIs, browser tools, automation and small products that can be tested quickly.

Build.Automate.Simplify.