Portfolio 2026 × 100% Lokesh Bhatia Product Designer

Hi, I’m Lokesh.

Product Designer solvingcomplex problems acrossfintech, AI & climate.

End-to-end design ownership across a trading app serving 200K+ users and 97% of company revenue, an AI agent platform taken from zero to public launch at $1M ARR in a month, and a carbon accounting product built up from rough wireframes.

See what I’ve built

Markets — move your cursor to steady themFig. 01

✦

portrait · 4:5

Selected work

Deep cuts, not a tour. Each one starts with the constraint and ends with what shipped — including the parts that didn't work.

Coming soon

Designing an agent that shows its work

0→1 AI product, concept to public launch

Case study in progress

What I do

Three things I'd bring on day one

Each one learned somewhere specific — the evidence is above.

Designing for AI that gets things wrong

Agent transparency, human-in-the-loop review, and recovery paths — shipped at Runable and ZZAZZ. When the model can be wrong, the interface has to carry it.

Getting fluent in hard domains, fast

Scope 3 emissions, F&O trading, brokerage revenue share, content pricing — four regulated domains in three years. I learn the business before I design the screen.

Owning a product end to end, solo

0→1 to public launch as the only designer, a trading app at 200K+ registered users, and a B2B platform built from nothing. Discovery, IA, interaction, and the design system underneath.

Impact

Numbers are the least interesting part of design work — but they're the part that survives the retro. Four I'd defend in a room, each with its receipt attached.

2.4×

Lead-to-client conversion in the two months after Rise Portal launched.

source: business team, partner dashboard

48%

Fewer partner support tickets once earnings and lead status went self-serve.

source: support team, relationship managers

97%

Of company revenue ran through the trading app I owned end to end.

source: rupeezy revenue reporting

~$1M

Annualised run-rate in Runable's first month post-launch — 0→1, no PM.

source: founder, revenue dashboard

Designing for AI

Models are confident, wrong, and fast. Three patterns I reach for when the interface can't promise a right answer — each one shipped, and each one rebuilt here so you can try it.

{{ demoLabel }} interactive — try it

Make undo cheaper than being right

A model that’s wrong one time in ten is unusable if fixing it costs a support ticket.

ai draft · to priya{{ draftBadge }}

Nothing commits until the user says so. Draft is the default state, not the exception — so being wrong costs a click, not a support ticket.

shipped at: runable · zzazz

An agent that shows its work

A spinner hides whether the system is working, stuck, or wrong. Narrating steps turns waiting into trust.

agent run · tap a step

"Plan a launch page for a pottery studio."

Waiting with narration beats waiting in the dark — every step is a claim the user can check, and every decision has a place to be wrong.

shipped at: runable ai

The first session is the product

AI tools get demoed, not adopted. People arrive curious and leave without shipping anything.

first session · tap a step{{ progLabel }}

AI tools get demoed, not adopted. I design the first real output a person gets — the feature list can wait for session two.

shipped at: runable ai

If your product has a model in it, these are the problems I've already made mistakes on.

Work experience

Sustainability, then fintech, now AI. Often the only designer on the product — discovery through design system.

You’ve seen the work.Let’s talk.

Open to full-time product design roles — remote, or Bengaluru / Noida / Gurgaon. Available immediately.

hi.lokeshux@gmail.com

I reply within a day.{{ copyAnnounce }}