In production
HX Labs
HumanX's internal AI product arm. I own product design, front-end implementation, back-end wiring, and product scoping — the whole delivery path, one person.
Taste that keeps working after I leave the room.
I design products, then build the rules, tools, and software that stop teams solving the same problem twice — design system to production React to back-end wiring.
Design builder / Systems thinker / AI-native
Rishabh Kabra
Selected work
In production
JuiceLabs AI
An AI creative platform serving enterprise consumer brands. Nine months, three roles' worth of arc: QC intern → design-system owner → the designer whose architecture core design now routes through.
Shipped · Team-adopted
AI-native Storybook
The design system rebuilt as an environment for agents, not just humans — so coding agents ship approved, published components instead of rebuilding their own.
Method, open-sourced ↗
Method · Demo on request
Image-to-UI compilation engine
An engine that turns screenshots and generated concepts into editable, auto-layout Figma builds — hardened with 206 regression tests across real product screens.
Live
Canon
A gate that catches stale owners, dates, and expired assumptions before an AI agent acts on them. Every verdict is computed from a live fact table, not painted on.
Open Canon ↗
Private system · Daily use
Mind Palace
A local AI pipeline on an 8GB laptop: a research radar that has scored 2,300+ videos and articles through a local model, at zero cloud cost.
The second time, something else does it.
My own repeated judgment, turned into installable skills my agents run.
Decision Compiler
Product decisions run through a bounded adversarial protocol: independent lenses, forced merges, a reviewer that grades the author. No premise settles unchallenged.
Review Fan-out
Large audits split across parallel reader agents, then an adversarial verifier that tries to refute their claims before anything is filed.
Pre-flight Council
Five advisors stress-test a plan before code is written; a calibration log scores past verdicts.
Image Megaprompt ↗
Turns a visual idea into a prompt that renders it the same way twice, in one frozen format.
Storybook Method ↗
The AI-native design-system operating loop itself, product-agnostic, reusable on any system.
Seventeen skills and counting — Figma reconstruction, image delivery contracts, design debates, stakeholder drafts, morning operations. Any judgment I repeat becomes one.
I left the room. It kept working.
Three things I built became other people's default way of working.
The org's workflow
Leadership watched the design-in-code pipeline run, then made it how the company works.
Designers shipping code
Non-technical designers raise and merge their own pull requests, off a git guide I wrote.
A system agents read
The design system publishes itself to coding agents, so the team ships from it without me in the loop.
I work beyond Figma.
Self-taught product designer and systems thinker, first-principles by habit. Product strategy, interaction design, design systems, AI-assisted workflows, and enough production code to carry the decision into reality.
Currently — JuiceLabs AI + HumanX (HX Labs) · Studied B.Tech CSE, 2021–2025
Let's build the thing behind the work.
If your team is solving the same product decision repeatedly, I'm interested.
