Mission-ready

Product Software
Engineer

Hi, I'm Ihor — I turn ideas into reliable, scalable products — from user experience and architecture through delivery.

View projectsLinkedInEmailGitHub

give it a try

Illustrated portrait of Ihor Levchenko

01 / About

I turn complex ideas into products that feel simple.

I’m a product-minded software engineer with a strong frontend foundation and ten years of experience. I work across UX, architecture, and delivery — turning ambiguous requirements into dependable products. I’ve built software for cybersecurity, payments, e-commerce, and data platforms, while also creating products of my own, including DoneHero.

02 / Projects

Selected work

Productivity2026

DoneHero

Goals, tasks, and habits turned into XP, levels, and self-defined rewards — a productivity system that pays you back.

Visit product
DoneHero productivity application
Product2025

Meridian notes

Notes, relational databases, whiteboards, and a daily journal in one markdown-native workspace.

Visit product
Meridian notes application
Trading system2022

Cbot

An adaptive trading bot combining machine learning, technical strategies, indicators, and stop-loss risk controls.

Cbot trading system

Productivity2026/ Case study

DoneHero

Making progress visible enough to sustain motivation.

DoneHero productivity application

Problem

Motivation fades when long-term goals feel distant, daily effort is hard to see, and conventional task or habit tools offer little reward beyond checking a box.

Product hypothesis

Turning effort into visible progress — XP, levels, scores, and self-defined rewards — can make small daily actions feel meaningful and keep users moving toward larger goals.

My role

Product strategy, UX/UI, frontend, backend, database design, and deployment.

Key design decisions

  • Goal → Epic → Task hierarchy that breaks big ambitions into shippable pieces.
  • Difficulty-based XP so effort, not busywork, drives progress.
  • Levels, scores, streaks, insights, and achievements that make progress visible at several time scales.
  • A unified Today view that connects each task and habit to a clear, immediate sense of progress.
  • Self-defined rewards users can redeem, linking completed effort to personally meaningful motivation.
  • An Inbox for frictionless quick capture.

Technical challenges

  • Balancing the XP and reward economy so progress feels attainable without making the system easy to game.
  • Keeping task, habit, and goal progress consistent across every view.
  • Building responsive, information-dense interfaces that stay calm.
  • Modeling recurring habits and streaks reliably.
  • Implementing it end to end with React, TypeScript, Fastify, and PostgreSQL.

Validation & iteration

Ran clickable prototypes past target users, folded in usability feedback each round, and cut ideas that tested poorly — an early points-only model was scrapped once users said streak pressure felt like the trackers they were escaping.

Outcome

The resulting system pairs every completed action with visible progress and lets users turn that progress into rewards they chose themselves, replacing guilt-driven productivity with a clearer feedback loop.

Product2025/ Case study

Meridian notes

Giving every thought the right shape without splitting work across apps.

Meridian notes application

Problem

Documents handle prose, spreadsheets handle structure, and canvas tools handle spatial thinking, but keeping them in separate apps fragments context and leaves related work out of sync.

Product hypothesis

A single workspace can support different modes of thinking without lock-in by connecting markdown notes, relational databases, an infinite canvas, and date-addressed daily notes.

My role

Product strategy, UX/UI, application architecture, frontend, backend, database design, and deployment.

Key design decisions

  • Built four connected content shapes: markdown notes, relational databases, infinite whiteboards, and one daily note per date.
  • Made every database row a full note, combining typed properties with space for the reasoning behind the row.
  • Linked notes, databases, and boards through @-mentions, relations, and live note references on the canvas.
  • Kept navigation fast with nested folders, per-item icons, a command palette, and a quick file switcher.
  • Used markdown for storage and import/export so users can move their work without a proprietary archive.

Technical challenges

  • Modeling notes and database rows as the same underlying content while supporting typed columns and multiple saved views.
  • Keeping filters, sorts, grouping, column settings, and table, board, list, and gallery layouts consistent.
  • Embedding live workspace references in notes and on an interactive, pannable canvas.
  • Preserving nested folder structure through markdown zip imports and full-workspace exports.
  • Delivering autosave and responsive navigation across a feature-dense web application.

Validation & iteration

The product was shaped around complete workflows: importing an existing markdown library, turning notes into structured views, linking research to a board, capturing a daily log, and exporting the workspace with its folders intact.

Outcome

Meridian now brings four modes of thinking into one searchable workspace, with unlimited notes, databases, and whiteboards available during early access and full markdown export to avoid lock-in.

Trading system2022/ Case study

Cbot

Building a trading system that learns while keeping risk bounded.

Cbot trading system

Problem

Fixed trading rules can become less effective as market behavior changes, while unconstrained automated decisions can turn a weak signal into an outsized loss.

Product hypothesis

A bot can make more disciplined decisions by combining machine-learning signals with explicit strategies, technical indicators, and non-negotiable risk controls.

My role

System design, strategy research, machine-learning experimentation, implementation, and risk-control design.

Key design decisions

  • Combined machine-learning outputs with rule-based strategies instead of relying on a single prediction source.
  • Used technical indicators as measurable inputs for entries, exits, and market-state assessment.
  • Applied stop-loss rules as a hard risk boundary independent of the strategy signal.
  • Recorded strategy performance so the system could compare results and improve its decision policy over time.

Technical challenges

  • Preparing market data and features without leaking future information into model training.
  • Coordinating signals from strategies, indicators, and machine-learning models when they disagree.
  • Separating strategy optimization from risk controls so self-improvement cannot remove safety boundaries.
  • Evaluating changes across different market conditions rather than overfitting to one period.

Validation & iteration

Strategies were evaluated against historical market periods with indicators, model signals, and stop-loss behavior assessed separately before being combined.

Outcome

Cbot established a modular foundation for testing trading ideas, enforcing downside limits, and feeding observed strategy performance back into future iterations.

03 / Arcade

Crashball

WebGL · three.jsSingle player~2 min

A 3D tribute to the first level of Crash Bash

Crashball is the opening arena of Crash Bash — four hovering bumper cars, one goal each, and a set of steel balls that get quicker the longer you last. You hold the south goal. Every ball that slips past costs a point, zero puts you out, and the last car standing takes the round.

  • 15 points each. A ball through your goal takes one away.
  • Corner cannons fire the balls in — watch for the flashing floor chevron.
  • Extra kick shoves nearby balls away — same as the original's square button.
  • Three round wins earns the trophy and clears the level.
Loads on demand · opens full screen

Loading the arena…

04 / Get in touch

Let's build something
worth shipping.

Open to product engineering roles and select freelance work. I usually reply within a day.

LinkedInEmail