Personal project — Independent · Google PlayProduct Design · Front-end · UI/UX

HeavyMuscle

A gym app with an AI coach that writes the plan and a logging screen built to be used standing up, with no signal.

What I can showLive on Google Play — install it and you are looking at the work.

HeavyMuscle: a five-day split with muscle recovery above the exercise list, and a live session logging weight by reps set after set — The plan, and the screen you actually use in the gym

Role: UI/UX and front-end
Team: Small team
Timeline: Ongoing
Platforms: Android, Wear OS
Stack: Figma, TypeScript, React Native

01Context

Same arrangement as Remote Work: I came in once it was running, improved the UI and UX, and worked on the front-end.

An AI coach builds a plan from one sentence and adjusts each session to your injuries, equipment and recovery. It shows the plates to load, generates warm-up sets, pre-fills every set with your last numbers and detects PRs. 873 animated exercises, six languages, fully offline.

02What I owned

Scope, in my own words.

Designed the screens before building them

Home, the empty state, both coach modes and the exercise library were drawn as one set in Figma. The empty state got a frame of its own — the first session is the one most people never start.

Figma frames for HeavyMuscle: the workout home, the empty state offering a quick start, the coach conversation, the coach builder thread, and the exercise library — Home, empty state, both coach modes, library

Improved the UI and UX

The home screen answers one question — what am I training today — before anything else. Muscle recovery sits above the exercise list, so the plan and the reason for it are read together.

HeavyMuscle training day: a five-day split with the current day expanded, muscle recovery per muscle group, and each exercise carrying its sets, reps, rest and a coaching note — Recovery state sits above the work it explains

Gave the AI coach two modes

Conversation for questions, Builder for a plan you can review and save. A chat that silently rewrites your training week is not something you want to discover mid-session.

HeavyMuscle AI coach in builder mode: a request for a knee-friendly session, and the generated plan listed day by day with sets and reps, waiting to be saved — Builder mode — the plan is proposed, then saved

Built the components that did not exist

No component layer, so every screen had been built on its own. I designed the components, implemented them and rebuilt the screens on top of them. It is offline-first, so it stays usable in a gym with no signal.

HeavyMuscle exercise library: a searchable list grouped by difficulty, each exercise tagged with its muscle group and the equipment it needs — The library, filtered by equipment and level

03The system

The screen that runs forty times a session

Logging a set happens constantly, standing up, often with no signal. Every control is thumb-sized and pre-filled with your last numbers, so the common case is two taps.

Time elapsed, sets left and estimated calories stay pinned while you scroll.

HeavyMuscle live session: elapsed and remaining time, sets completed and calories pinned at the top, each set logged as weight by reps with a confirm circle, and swap, skip, rest and notes controls under the exercise — Mid-session — weight by reps, one tap to confirm

04Outcome

What changed.

873: Animated exercises
6: Languages supported
5.0: Current rating

Newer than Remote Work and currently rated 5.0. There is a Wear OS app too, so a session can be run from the watch.

Next — Role

CoinFello

Built the design system and implemented it in code. A DeFi product with an AI agent at the centre of it.