Designing AI-native, agentic experiences
Cross-domain multi-agent frameworks, agentic analytics, and explainable human-in-the-loop systems — turning complex, regulated data into products a person can actually decide from. Held together by design systems I build and document, alongside PM and engineering.
- Experience
- 5+ years, 0→1
- Domains
- AI · Fintech · Logistics
- KYC & onboarding
- 95% faster
- Agent memory
- Multi-domain context
A low-key dream of mine: to join — and one day nurture — a team that cares this deeply about craft, quality, and building great tools.
Selected work
PrivateOptaire AI Fabric
Designing the Intelligence Layer for Enterprise Workflows
PrivateOptaire Business Intelligence
Designing an AI-Native Decision Intelligence Experience
PrivateOptaire Design System
Building the Foundation for a Multi-Product Ecosystem
PrivateLogistics Application
Replacing paper registers and phone calls with a real-time operations platform for a multi-factory paddy procurement and rice distribution business.
PrivateSearch Suggestions UX in an AI Chatbot
Performance-Driven Search Suggestion Experience for Sales Teams
PrivateHow Cookies Improved Referral-Based Onboarding in Tara Connect
Used cookies to preserve referral intent across sessions and reduce onboarding drop-offs.
PrivateRecovered 30% More Users by Fixing a Hidden Onboarding Friction
A systems-driven redesign of Entera's email verification flow that reduced cognitive load, guided user attention, and significantly improved activation.
PrivateInformed Investor: Stock Market Learning App UI & Visual Design
Designed and built the full product from scratch. This case study focuses on the UI design, visual clarity, and screen craft.
PrivateZentra: Bank Statement Analyser for Financial Institutions
Turning raw bank statements into clear, defensible lending decisions — decision intelligence for lenders.
Icon Foundry
View on GitHub (opens in a new tab)A design-system-native icon foundry for creating, validating and publishing custom icons in your own visual language. A team writes down how its icons are built; the foundry draws every new icon from those rules, checks it against them, ships the set to code and syncs it into Figma.
Fig · One brief, compiled — the engine's real output, not a drawing of it. Technical v0.2.0 · 4ede017
01 · IntentA sentence in. The keyword parser finds what it is made of — no model needed.
Brief
“a cold-storage warehouse”
Parsed by keywords · no model
subject · warehousemodifier · snowflake
The set · 14 icons · shown at 2×
1 of 6
Build log
Still not true
A team builds a real set with it
Every threshold in the audit is still supported by synthetic evidence alone — the largest gap in the project.
Still not true
Run the plugin inside real Figma
The one thing no test can stand in for.
M22–M26 · Construction
The construction editor
Straight segments on the angle set, then rounded joins — how a shape is built becomes the language's, and editable.
M19–M21 · Filled
Two styles
The language owns the filled policy, a filled version is a variant with a status, and the set says which icons lack one.
M17–M18 · The screen
One page for the language
Everything that changes the drawing beside a live canvas of every part, and a panel that moves traits from plain words.
Playground
Smaller things built to answer a question rather than to ship — a plugin, a demo, a tool, each shown doing the thing it does.
still here??
Things I love doing
- Drawing the thing before arguing about it
- Naming things until the name does the explaining
- Reading the logs with the engineers
- Sitting in on support calls
- Building the tool instead of filing the ticket
- Throwing away the first three versions
- Teaching someone the shortcut I just found
