mono 12 · tracking 0.1em · uppercase Fraunces 300 · clamp(48→92px) · ls −0.03em · #EDEBE6 italic → var(--accent) · #A8956E Syne 18 / 1.7 · #87847C · 560px measure Fraunces 32 · tabular-nums — every value traceable to Tokens/Colors.md
Lead Product Designer · Ex-D.E.Shaw · IIT Bombay MDes

I design with AI,
not around it.

AI is my daily creative partner. I prototype in code, and I work in problem spaces without established conventions — sole designer across 4 products on a financial analytics platform for institutional investors.

4
products shipped
1
designer
200+
screens
50K+
notes in prod
the loop, below
The Loop
My design system talks to AI.
Watch it run.

This is the real workflow behind every Arcana feature. AI reads a 300-note Obsidian vault of tokens, components, and patterns, resolves the exact values, and renders a working prototype. The file names, tokens, and component below are the actual ones.

standby
01 The Vault
Real files — hover one to trace it through the loop.
02 The Read
AI traverses tokens & anatomy on demand.
03 The Render
A working prototype, from the exact tokens read.
Arcana — Portfolio Grid · generated prototype
header #102136 · scroll #324F79 · rows 21px · Open Sans / Inter — traceable to the vault
80%
accurate on first pass
Minutes
idea → clickable prototype
7 in 7
prototypes in days — calendar sprint
8+
hedge funds on the system

The thumbnails in the case studies below are live prototypes this loop produced, and the graph further down is the actual vault. The same loop took the calendar feature to production handoff in seven days.

How I Work
Seven phases. One designer.
AI in every one.

AI doesn't replace the steps. It makes all of them possible at once, across four products, from the first client call to production handoff.

01
I'm on every client call: Longaeva, Aperture, Square Point, D1 Capital. The direction comes from hearing the problem myself rather than reading a summary of it. AI turns the call notes into structured problem statements and requirements, and the framework for what we build is set before any design starts.
02
Every feature starts with a live audit of four or five incumbents (Bipsync, Verity, EDS, AlphaSense, FactSet) across capability, UX, and positioning. AI speeds up the comparison matrices and surfaces the patterns across them, which shows where we can take ground the others have left open. Positioning feeds the product direction from the start.
03
I co-author the PRDs with the PM: user workflows, edge-case matrices, interaction specs, and the rationale, all before a screen is drawn. AI stress-tests the edge cases and fills in scenario coverage. Seven PRDs have come together this way; the largest ran to 378K characters. Most of the alignment gets settled here, well ahead of review.
04
This is where the design system earns its keep. AI reads the vault, 300 linked notes of tokens, components, and patterns, and builds interactive HTML prototypes from the real tokens. The first pass is usually close enough to react to, so decisions get tested in minutes instead of weeks. The calendar feature went through seven full prototype rounds in seven days.
Traditional: 2-3 weeks in Figma for one high-fidelity prototype. AI-assisted development workflows: multiple interactive prototypes per day, each clickable and stakeholder-ready.
05
Stakeholders click through working prototypes instead of annotated mockups. Feedback gets sharper when people react to real behavior: "too messy" turns into a specific interaction note. Each round comes back as an updated prototype within the hour, and decisions land faster because a working screen answers the questions a static one leaves open.
06
Handoff is a package engineers can build from: theme CSS with every token mapped, React components, configuration objects, and data-schema transforms. A developer runs npm install and the feature renders with the system already applied.
07
Every confirmed component goes into the vault with its full anatomy, token tables, and states. The system grows with each feature, so by month six a new one arrives most of the way built. The vocabulary and precedents are already there.
AI-Native Design Systems
The system the AI reads.

Most design systems are built for humans to browse. Mine is built for AI to consume and apply. 300 interconnected Obsidian notes form a knowledge graph that gets traversed in real time. Hover, drag, and zoom the graph below — this is the live structure, not an illustration.

Best experienced on desktop — drag, zoom, and hover the nodes.

Token traceability
Every hex value in every component traces back to a token file. The AI never invents colors. Consistency is enforced by structure.
Lazy loading
AI reads two files when a session starts and pulls component docs only when it needs them, so its attention stays on the design problem.
Compounding returns
Month 1: 3-4 rounds to accurate output. Month 6: right on the first pass. The vault grows. Every session starts further ahead.
Read the full workflow →
Selected Work
Case studies.
Zero to one
Product direction, information architecture, and interaction framework defined before a single screen exists.
Systems AI can read
300 interconnected nodes, structured for machine consumption. The loop above runs on it.
Workflow design
Research, decision support, and portfolio analytics in one place — built to end the analyst's tab-switching.
Data-dense craft
Interfaces where density and legibility are the product. Every hex traceable to a token.
CS-004 · ARCANA AI
Teaching an AI to Show Its Work
A trust & validation layer for AI-generated analytics — computation receipts, honest caveats, and edit-and-rerun validation so investors can act on AI-computed numbers.
0→1 AI feature 2 prototypes 5-part receipt
CS-002 · ARCANA
Building an Enterprise RMS from a Blank Canvas
Platform strategy and workflow architecture across 9 feature areas, solo, over 9 months. Full enterprise adoption at 8+ hedge funds — displacing Bipsync, Verity, and EDS.
200+ screens 8+ clients 9 months
CS-003 · ARCANA · DEEP DIVE
Where Research Meets the Market
Research intelligence platform unifying 33K+ notes with live financials. 5 views, inline estimates, and decision support — designed for zero context-switching.
5 views 10yr data 33K+ notes
CS-001 · ARCANA · DEEP DIVE
Zero to Production Handoff in 7 Days
Seven interactive prototypes to production handoff in seven days, on AI-assisted workflows and a codified design system.
7 prototypes 7 days

The shape of the work, in four numbers.

1
designer
3
incumbent tools replaced
300
design system nodes
378K
characters, largest PRD
Background
Resume.
Experience
Lead Product Designer
Aug 2025 - Present
Arcana · Full-time · Remote
Platform strategy and design systems lead across 4 products (web, iOS, AI, risk analytics) for a financial analytics platform serving institutional investors. Drove enterprise adoption at 8+ hedge funds with a full RMS that displaced Bipsync, Verity, and EDS. 200+ screens, 9 feature areas, 300-node AI-native design system. Co-authored 7 PRDs. Shipped a calendar feature via AI-assisted development workflows (7 prototypes in 7 days) with production React handoff. Established WCAG AA accessibility across all interfaces — high-contrast color system, keyboard-navigable data grids, screen-reader compatible financial tables.
Senior Interaction Designer
Jun 2023 - Aug 2025
D. E. Shaw & Co · Hyderabad
Design lead for the HR vertical. Enterprise talent management software with 4K+ daily users. Worked with CXO-level stakeholders on the firm's employee assessment tool. Managed 3 interns (all offered PPOs). Conducted 30+ hiring interviews across campus and market drives.
Interaction Designer
Jul 2021 - May 2023
D. E. Shaw & Co · Hyderabad
Design lead for the Options Trading Group. Contributed 50+ components to the Gravity design system. Built and maintained the in-house charting library (2.5M instances/week firm-wide). Designed the CI/CD tool that drove a 30% boost in developer productivity.
UX Design Intern
Jun - Jul 2020
Honeywell · Remote
Designed and shipped a GUI for the sensor configuration workflow for Honeywell's Asset Performance Management System. Secured a PPO.
Junior Designer
Aug 2018 - Jul 2019
Ghosh Bose & Associates · Kolkata
Design and branding for West Bengal government's flagship startup incubator project.
Education
IIT Bombay
Master of Design, Human Computer Interaction
2019 - 2021
Research Assistant at the Human Factors and Sociotechnical Systems Studio under Prof. Vivek Kant
NIT Patna
Bachelor of Architecture
2013 - 2018
Selected in the National Best Thesis Awards, 2018
Recognition & Writing
Your Design System Is Useless to AI
LinkedIn article, 2026 — argues traditional design systems are invisible to AI tools and proposes a machine-readable structure, grounded in a working implementation.
Read on LinkedIn →
Best Paper Award, INCOSE HSI 2021
Holistic Asset Performance Management Systems: Speculative Design for Industrial APM Systems and Human Work Experience in an Industry 5.0 Setting
Published in Wiley →
Skills
Design
Code Prototyping Interaction Design Design Systems Information Architecture User Research Accessibility (WCAG) Design Handoff
Tools
Figma Claude / LLMs Cursor HTML/CSS/JS Obsidian Blender Adobe CC
Domain
AI Interactions Conversational Interfaces Agentic Systems Data-Dense Interfaces Financial Analytics Enterprise Software