Open · taking briefs
London studio , ·
UX × Computational·v0.5
Session 0001·Interactions 0000·⌘Kcommands
Q41·001 · HERO
/OPENING
[a]
FILE 001/007
NOTE: I split my design personality into two characters because my background is split. Crux represents the architectural, structural thinking.
Leela is the rapid-prototyping, chaotic energy. Need to make sure the contrast between them is obvious from the first scroll.
Sketch note · HEROdraft 4
Why are there two characters on the homepage?
Most designers pick one style. I couldn't to I do careful UX research and wild architecture experiments. So instead of hiding it, I made two characters: Crux (the careful one) andLeela (the experimental one). Same person, two energies. That's the whole pitch.
◯ honest about itkept
01 · CRUX
focusedgroundedCRUX · POSITION 01
02 · LEELA
tilted fwdmomentumLEELA · POSITION 02
Portfolio · Two tracks · UX & Computational · 2019 to 2026

UnmeshLele

Architect trained Behavioural UX Designer & computational designer. Two tracks, one mind · product UX on screens, parametric systems in space.

UX case studies→Computational design
Scroll
Q41·002 · UX CASES
/TRACK·01
[b]
FILE 002/007
The standard 3-column grid was too safe and boring. Every portfolio on dribbble uses it. Instead, I want the user to scroll through immersive full width scenes.
Sketch note · UX WORKdraft 3
Why not just use normal project cards?
Tiny cards hide all the interesting stuff. I wanted each project to feel likewalking into a room to big letters in the background, the actual phone screen in front, real numbers floating beside it. You should feel the size of the problem, not just see a thumbnail.
◆ risky butworth it
Track 01 · Product & UX

The problem underneath the brief.

Three projects where the real problem only surfaced after we stopped answering the wrong one.

Featured casesFull case cards · auto slides every 5 seconds
View case study →
Case 01 / 03 · Direct Line Group · Motability
Insurance · FNOL
Jun 2025 to Jan 2026
01
THE PROBLEM UNDERNEATH · THE PROBLEM UNDERNEATH ·
Brieflies
What happened?One question at a timeCONTINUE
0102030405SEQUENCED · NOT SPLAYED
Claim loggedReference · FNO-19X42KNEXT STEPSAssessor calls within 24hCourtesy vehicle: arrangedYou're not alone in this
85%Error reduction in claim loggingUser tested · before and after usability task comparison
22%Admin time saved · 20% faster resolutionPrototype estimate · task time comparison

MobilityFlow

View case study →

Brief vs. realityThe brief said "fix the form." The real problem was cognitive overload during a trauma moment. People weren't failing the form to the form was failing them at the worst hour of their year. We redesigned the entire First Notice of Loss journey around sequenced attention, not splayed inputs.

Behavioural UXFNOLWCAG 2.2 AADesign to code
Project typeInsurance FNOL · accessibility first · client concept · mobile web
TimelineJun 2025 to Jan 2026
RoleResearch · UX · UI · prototype
PlatformMobile web
StackFigma · React prototype
StatusCase study ready
Evidence levelUser tested
My contributionLead UX/UI + prototype
Best artifactSequenced FNOL flow
Outcome85% claim logging error reduction
Case authorisation UX-01-FNL-MFLO · Category insurance / claims · Issue 2025-Q4 · Lead · U. Lele
View case study →
View case study →
Case 02 / 03 · Fintech · Gov · AI
GDS Alpha passed 11/11
BJ Fogg model applied
02
TRUST IS EARNED NOT SHIPPED · TRUST IS EARNED ·
ModelFogg
Hi, I'm Billie.I'll start small.Ask me something simple
Is my deposit protected?Yes to up to £85k under FSCS.Here's where that number comes from →Show sourceExplain moreABILITY → MOTIVATION → TRIGGER
Based on your last 3 weeksYou typically overspend on Thursdays.Want me to flag it tomorrow morning?Yes, flag itNot this weekTRUST · EARNED0.84Earned, not shipped.
11/11GDS Alpha assessment passedMeasured · GDS checklist
31%Confidence lift in first use testingUser tested · first use sessions

Billie

View case study →

Brief vs. realityThe brief said "build an AI assistant."The real problem was trust. People don't want a smarter bot to they want a slower one that earns its next question. We used BJ Fogg's behaviour model to design the trust curve before the interface.

AI · ConversationalBehavioural modelGDS AlphaFintech · Gov
Project typeFintech AI · behavioural finance · side project · Gemini prototype
Timeline2025-Q2
RoleAI UX · behavioural modelling · prototype
PlatformConversational AI
StackFigma · Gemini prototype
StatusAlpha case study
Evidence levelPrototype estimate
My contributionProduct UX · trust model · interaction design
Best artifactTrust calibrated assistant flow
OutcomeGDS Alpha 11/11 criteria passed
Case authorisation UX-02-AI-BILLIE · Category conversational / fintech · Issue 2024-Q4 · Lead · U. Lele
View case study →
View case study →
Case 03 / 03 · Self initiated · React PWA
Identity-led gamification
Design to code
03
IDENTITY BEFORE OUTPUT · IDENTITY BEFORE OUTPUT ·
Loopdaily
TUE · 07:12Good morning,builder.Today's one thingWrite 400 words on Project Ulysses25 min · no phoneBegin ritual
STREAK · 42 DAYSYou show up. Daily.That's the identity, not the task.+1 TODAY
U.IDENTITY TIER · 03The BuilderUnlocked at day 42NEXTThe Shipper · 90 days
42dDay streak · median power userSynthetic/demo data · power user model
3×Retention vs. task list baselinePrototype estimate · benchmark model

Unwind

View case study →

Brief vs. realityThe brief said "productivity app." The real problem was identity. People don't want to be productive to they want to feel like someone who already is. We built the loop around identity reinforcement, not task completion. Shipped React PWA, full design to code.

React · PWAIdentity loopsDesign to codeSelf initiated
Project typeProductivity OS · identity loops · self initiated React PWA
Timeline2024
RoleResearch · UX/UI · React build
PlatformProgressive web app
StackReact · PWA · design system
StatusLive prototype
Evidence levelSynthetic/demo data
My contributionSolo product design + frontend build
Best artifactIdentity loop onboarding prototype
Outcome42d median power user streak
Case authorisation UX-03-PWA-UNWND · Category productivity / habit · Issue 2024 · Lead · U. Lele
View case study →
Q41·003 · PRINCIPLE
/INTERLUDE
[c]
FILE 003/007
I cut this section in half. The long version was a whole essay. Keeping it to just one punchy statement about how I approach design briefs makes it hit harder.
Sketch note · BIG QUOTEdraft 3
The one sentence everything hangs on:
Rewrote this three times. Cut all the fancy words. What's left is the simple truth: what someone asks for is never exactly what they actually need. My job is to find the gap. The whole site builds up to this line, then proves it afterwards.
▲ core beliefkept
, A working principle ,

Every brief lies.

Not maliciously to just incompletely. The brief describes what someone thinks they need. My first job is to find what's underneath it.

Q41·004 · COMP
/TRACK·02
[d]
FILE 004/007
This interactive canvas isn't a gimmick to it's the core thesis of my parametric work. Every project below is just a static screenshot of one configuration of this tool.
Sketch note · STUDIOdraft 3
Why a live playground instead of static images?
Architecture portfolios are always just pretty pictures to render after render after render. Dead. But when you're actually making this kind of work, it moves and responds. So I put the real tool at the top. Each project below is just a different setting on this instrument.
◆ hardest decisionon the page
Track 02 · Computational Design

Before pixels, I designed buildings.

Four years of parametric architecture to competition entries, thesis work, built homes , ending in 2023 when the medium moved from buildings to screens. The spatial thinking came with it.

Parametric studio · live

Every project below is one configuration of this.

Parametric design is a system with dials. Every building I made was a different setting. Move your cursor on the canvas. Switch presets. Pull the dials yourself.

Zaltt Labs · Parametric builder
Ed. 01 ·  2026  ·  v0.3
Parametric Grid AttractorRectilinear

Drag the canvas to shape the field. Save Art PNG downloads the current artwork.

, Six real configurations · ground truth projects · 2019 to 2023 ,
Architecture project matrix

A live planning board. Cards recompose into full screen massing studies, revealing new adjacencies between studio, thesis, built work, and experiments without overlapping the plan.

Field plan
Q41·005 · LABS
/TRACK·03
[e]
FILE 005/007
This is the messy sandbox where next year's polished case studies are currently being drafted and tested in public.
Sketch note · LABSdraft 2
Why have an experiments section?
Client work has to be polished. This doesn't. Labs is where half-baked ideasgo to be tested out loud to before they're good enough for a real project. Most of them fail, and that's the whole point. Failure in public = learning faster.
◯ always runninged.01
Zaltt Labs · Active

Where I test in public.

Labs is where I publish unfinished thinking. No case studies, no polish. Just the actual experiments, honestly documented.

EXP · 001AI for heritage architectureCan a model understand what a buildingmeans before suggesting what to repair? Active build; beachhead = UK restoration architects.
EXP · 002The threshold of commitmentMapping the exact moment a user decides. Behavioural UX applied to conversion micro moments.
EXP · 003The anti brief methodA systematic way to invert a client brief and surface the problem it's hiding.
Q41·006 · ABOUT
/CONTEXT
[f]
FILE 006/007
I kept the facts panel extremely tight. One line each. No bio paragraph bloat. Recruiters just want to scan the keywords anyway.
Sketch note · ABOUTdraft 2
This isn't a life story:
Recruiters scan About sections in 8 seconds. So the right side is pure facts to role, school, tools, what makes me different. The left side? Just the one paragraph that actually matters. No fluff. If you can't say it in 8 seconds, you don't know what you do.
◯ design rule8 sec scan
About

I'm Unmesh.

I'm an architect who moved into UX. I treat interfaces like buildings , load bearing decisions, honest materials, and a belief that every brief is hiding the real problem somewhere underneath it.

Before pixels, I spent six years on parametric architecture to competition entries in Mannheim, a thesis on urban memory, a family home with a jali screen that's still standing. Then an MSc in Human Computer Interaction at Nottingham shifted the medium, not the method.

Now I design behavioural UX: insurance journeys where people are in actual distress, AI assistants that have to earn trust before being useful, productivity tools where the real feature is identity. I also build what I design to React, PWAs, full design to code when a team needs it.

Zaltt is my studio's name to for now it's just the domain this portfolio lives on. A fuller thing, eventually.

Unmesh Lele · Designer
Read full About pageView Product/UX case studiesView computational design case studies
Q41·007 · CONTACT
/END
[g]
FILE 007/007
I killed the complex contact form. Opening the user's native mailto app respects their time and feels much more personal.
Sketch note · CONTACTdraft 3
Why not just put an email address?
The clickable chips at the top do the hard work to they sort what you want before you type anything. "Hiring" starts a different conversation than "Heritage AI". And the send button? It just opens your own email app. No database, no tracking. Respects your time.
◆ simple on purposekept

Let's talk.

Role conversations, consulting briefs, heritage AI collaborations, or "what's Leela up to this week." I reply within a day.

Opens your mail app · unmeshux@gmail.com · No tracking, no forms service middleman
LinkedInBehanceUX CVCOMPUTATIONAL CVRead Labs
Open · taking briefs
London studio , ·
UX × Computational·v0.5
Session 0001·Interactions 0000·⌘Kcommands
Q41·001 · HERO
/OPENING
[a]
FILE 001/007
NOTE: I split my design personality into two characters because my background is split. Crux represents the architectural, structural thinking.
Leela is the rapid-prototyping, chaotic energy. Need to make sure the contrast between them is obvious from the first scroll.
Sketch note · HEROdraft 4
Why are there two characters on the homepage?
Most designers pick one style. I couldn't to I do careful UX research and wild architecture experiments. So instead of hiding it, I made two characters: Crux (the careful one) andLeela (the experimental one). Same person, two energies. That's the whole pitch.
◯ honest about itkept
01 · CRUX
focusedgroundedCRUX · POSITION 01
02 · LEELA
tilted fwdmomentumLEELA · POSITION 02
Portfolio · Two tracks · UX & Computational · 2019 to 2026

UnmeshLele

Architect trained Behavioural UX Designer & computational designer. Two tracks, one mind · product UX on screens, parametric systems in space.

UX case studies→Computational design
Scroll
Q41·002 · UX CASES
/TRACK·01
[b]
FILE 002/007
The standard 3-column grid was too safe and boring. Every portfolio on dribbble uses it. Instead, I want the user to scroll through immersive full width scenes.
Sketch note · UX WORKdraft 3
Why not just use normal project cards?
Tiny cards hide all the interesting stuff. I wanted each project to feel likewalking into a room to big letters in the background, the actual phone screen in front, real numbers floating beside it. You should feel the size of the problem, not just see a thumbnail.
◆ risky butworth it
Track 01 · Product & UX

The problem underneath the brief.

Three projects where the real problem only surfaced after we stopped answering the wrong one.

Featured casesFull case cards · auto slides every 5 seconds
View case study →
Case 01 / 03 · Direct Line Group · Motability
Insurance · FNOL
Jun 2025 to Jan 2026
01
THE PROBLEM UNDERNEATH · THE PROBLEM UNDERNEATH ·
Brieflies
What happened?One question at a timeCONTINUE
0102030405SEQUENCED · NOT SPLAYED
Claim loggedReference · FNO-19X42KNEXT STEPSAssessor calls within 24hCourtesy vehicle: arrangedYou're not alone in this
85%Error reduction in claim loggingUser tested · before and after usability task comparison
22%Admin time saved · 20% faster resolutionPrototype estimate · task time comparison

MobilityFlow

View case study →

Brief vs. realityThe brief said "fix the form." The real problem was cognitive overload during a trauma moment. People weren't failing the form to the form was failing them at the worst hour of their year. We redesigned the entire First Notice of Loss journey around sequenced attention, not splayed inputs.

Behavioural UXFNOLWCAG 2.2 AADesign to code
Project typeInsurance FNOL · accessibility first · client concept · mobile web
TimelineJun 2025 to Jan 2026
RoleResearch · UX · UI · prototype
PlatformMobile web
StackFigma · React prototype
StatusCase study ready
Evidence levelUser tested
My contributionLead UX/UI + prototype
Best artifactSequenced FNOL flow
Outcome85% claim logging error reduction
Case authorisation UX-01-FNL-MFLO · Category insurance / claims · Issue 2025-Q4 · Lead · U. Lele
View case study →
View case study →
Case 02 / 03 · Fintech · Gov · AI
GDS Alpha passed 11/11
BJ Fogg model applied
02
TRUST IS EARNED NOT SHIPPED · TRUST IS EARNED ·
ModelFogg
Hi, I'm Billie.I'll start small.Ask me something simple
Is my deposit protected?Yes to up to £85k under FSCS.Here's where that number comes from →Show sourceExplain moreABILITY → MOTIVATION → TRIGGER
Based on your last 3 weeksYou typically overspend on Thursdays.Want me to flag it tomorrow morning?Yes, flag itNot this weekTRUST · EARNED0.84Earned, not shipped.
11/11GDS Alpha assessment passedMeasured · GDS checklist
31%Confidence lift in first use testingUser tested · first use sessions

Billie

View case study →

Brief vs. realityThe brief said "build an AI assistant."The real problem was trust. People don't want a smarter bot to they want a slower one that earns its next question. We used BJ Fogg's behaviour model to design the trust curve before the interface.

AI · ConversationalBehavioural modelGDS AlphaFintech · Gov
Project typeFintech AI · behavioural finance · side project · Gemini prototype
Timeline2025-Q2
RoleAI UX · behavioural modelling · prototype
PlatformConversational AI
StackFigma · Gemini prototype
StatusAlpha case study
Evidence levelPrototype estimate
My contributionProduct UX · trust model · interaction design
Best artifactTrust calibrated assistant flow
OutcomeGDS Alpha 11/11 criteria passed
Case authorisation UX-02-AI-BILLIE · Category conversational / fintech · Issue 2024-Q4 · Lead · U. Lele
View case study →
View case study →
Case 03 / 03 · Self initiated · React PWA
Identity-led gamification
Design to code
03
IDENTITY BEFORE OUTPUT · IDENTITY BEFORE OUTPUT ·
Loopdaily
TUE · 07:12Good morning,builder.Today's one thingWrite 400 words on Project Ulysses25 min · no phoneBegin ritual
STREAK · 42 DAYSYou show up. Daily.That's the identity, not the task.+1 TODAY
U.IDENTITY TIER · 03The BuilderUnlocked at day 42NEXTThe Shipper · 90 days
42dDay streak · median power userSynthetic/demo data · power user model
3×Retention vs. task list baselinePrototype estimate · benchmark model

Unwind

View case study →

Brief vs. realityThe brief said "productivity app." The real problem was identity. People don't want to be productive to they want to feel like someone who already is. We built the loop around identity reinforcement, not task completion. Shipped React PWA, full design to code.

React · PWAIdentity loopsDesign to codeSelf initiated
Project typeProductivity OS · identity loops · self initiated React PWA
Timeline2024
RoleResearch · UX/UI · React build
PlatformProgressive web app
StackReact · PWA · design system
StatusLive prototype
Evidence levelSynthetic/demo data
My contributionSolo product design + frontend build
Best artifactIdentity loop onboarding prototype
Outcome42d median power user streak
Case authorisation UX-03-PWA-UNWND · Category productivity / habit · Issue 2024 · Lead · U. Lele
View case study →
Q41·003 · PRINCIPLE
/INTERLUDE
[c]
FILE 003/007
I cut this section in half. The long version was a whole essay. Keeping it to just one punchy statement about how I approach design briefs makes it hit harder.
Sketch note · BIG QUOTEdraft 3
The one sentence everything hangs on:
Rewrote this three times. Cut all the fancy words. What's left is the simple truth: what someone asks for is never exactly what they actually need. My job is to find the gap. The whole site builds up to this line, then proves it afterwards.
▲ core beliefkept
, A working principle ,

Every brief lies.

Not maliciously to just incompletely. The brief describes what someone thinks they need. My first job is to find what's underneath it.

Q41·004 · COMP
/TRACK·02
[d]
FILE 004/007
This interactive canvas isn't a gimmick to it's the core thesis of my parametric work. Every project below is just a static screenshot of one configuration of this tool.
Sketch note · STUDIOdraft 3
Why a live playground instead of static images?
Architecture portfolios are always just pretty pictures to render after render after render. Dead. But when you're actually making this kind of work, it moves and responds. So I put the real tool at the top. Each project below is just a different setting on this instrument.
◆ hardest decisionon the page
Track 02 · Computational Design

Before pixels, I designed buildings.

Four years of parametric architecture to competition entries, thesis work, built homes , ending in 2023 when the medium moved from buildings to screens. The spatial thinking came with it.

Parametric studio · live

Every project below is one configuration of this.

Parametric design is a system with dials. Every building I made was a different setting. Move your cursor on the canvas. Switch presets. Pull the dials yourself.

Zaltt Labs · Parametric builder
Ed. 01 ·  2026  ·  v0.3
Parametric Grid AttractorRectilinear

Drag the canvas to shape the field. Save Art PNG downloads the current artwork.

, Six real configurations · ground truth projects · 2019 to 2023 ,
Architecture project matrix

A live planning board. Cards recompose into full screen massing studies, revealing new adjacencies between studio, thesis, built work, and experiments without overlapping the plan.

Field plan
Q41·005 · LABS
/TRACK·03
[e]
FILE 005/007
This is the messy sandbox where next year's polished case studies are currently being drafted and tested in public.
Sketch note · LABSdraft 2
Why have an experiments section?
Client work has to be polished. This doesn't. Labs is where half-baked ideasgo to be tested out loud to before they're good enough for a real project. Most of them fail, and that's the whole point. Failure in public = learning faster.
◯ always runninged.01
Zaltt Labs · Active

Where I test in public.

Labs is where I publish unfinished thinking. No case studies, no polish. Just the actual experiments, honestly documented.

EXP · 001AI for heritage architectureCan a model understand what a buildingmeans before suggesting what to repair? Active build; beachhead = UK restoration architects.
EXP · 002The threshold of commitmentMapping the exact moment a user decides. Behavioural UX applied to conversion micro moments.
EXP · 003The anti brief methodA systematic way to invert a client brief and surface the problem it's hiding.
Q41·006 · ABOUT
/CONTEXT
[f]
FILE 006/007
I kept the facts panel extremely tight. One line each. No bio paragraph bloat. Recruiters just want to scan the keywords anyway.
Sketch note · ABOUTdraft 2
This isn't a life story:
Recruiters scan About sections in 8 seconds. So the right side is pure facts to role, school, tools, what makes me different. The left side? Just the one paragraph that actually matters. No fluff. If you can't say it in 8 seconds, you don't know what you do.
◯ design rule8 sec scan
About

I'm Unmesh.

I'm an architect who moved into UX. I treat interfaces like buildings , load bearing decisions, honest materials, and a belief that every brief is hiding the real problem somewhere underneath it.

Before pixels, I spent six years on parametric architecture to competition entries in Mannheim, a thesis on urban memory, a family home with a jali screen that's still standing. Then an MSc in Human Computer Interaction at Nottingham shifted the medium, not the method.

Now I design behavioural UX: insurance journeys where people are in actual distress, AI assistants that have to earn trust before being useful, productivity tools where the real feature is identity. I also build what I design to React, PWAs, full design to code when a team needs it.

Zaltt is my studio's name to for now it's just the domain this portfolio lives on. A fuller thing, eventually.

Unmesh Lele · Designer
Read full About pageView Product/UX case studiesView computational design case studies
Q41·007 · CONTACT
/END
[g]
FILE 007/007
I killed the complex contact form. Opening the user's native mailto app respects their time and feels much more personal.
Sketch note · CONTACTdraft 3
Why not just put an email address?
The clickable chips at the top do the hard work to they sort what you want before you type anything. "Hiring" starts a different conversation than "Heritage AI". And the send button? It just opens your own email app. No database, no tracking. Respects your time.
◆ simple on purposekept

Let's talk.

Role conversations, consulting briefs, heritage AI collaborations, or "what's Leela up to this week." I reply within a day.

Opens your mail app · unmeshux@gmail.com · No tracking, no forms service middleman
LinkedInBehanceUX CVCOMPUTATIONAL CVRead Labs