AI Product Designer · 15 Years · Data-Intensive Systems

Making AI interfaces people actually understand.

I'm Arpit Maheshwari — a product designer specialising in AI and data-intensive products. Based in Indore, India, working remotely across US, UK, and EU time zones. I design intelligent digital experiences where AI amplifies human capability — from smart dashboards and ML-powered screening tools to accessible, launch-ready platforms.

15
Years deep
8+
Industries
550k
Users impacted

About

Arpit Maheshwari — Product designer for AI and data-intensive products, based in Indore, India

Product designer for
AI & data-intensive products.

Most designers treat AI outputs as isolated UI problems. I think in systems — understanding how products evolve from standardized processes → sensing/monitoring → emergent intelligence — then design each layer accordingly.

I code, I understand ML models deeply, I architect design systems, and I mentor teams through scale. That rare combination is why teams bring me back for long-term partnerships. Based in Indore, India, working remotely across US, UK, and EU time zones.

AI-Native Thinking
Designing for ML outputs, algorithmic trust, and intelligent interfaces
Data-Fluent
Turning complex data and AI signals into clear, actionable UX
Accessibility-First
Universal design — including AI-powered products — for all users
Systems Thinker
Products that scale with both users and AI capabilities

Selected Work

Intelligent work
that ships.

I design products where AI does the heavy lifting and humans stay in control. Most work is under NDA — but each project below shows the challenge, my specific role, and measurable outcomes.↗ Connect on LinkedIn for detailed case walkthroughs
NDA Protected
AdTech · Programmatic Platform

Disrupting Outdoor Advertising with an AI Based Programmatic Platform

Designing an AI-driven programmatic platform that uses machine learning to aggregate inventory, automate campaign planning, and optimize media buying across the US and UK outdoor advertising market. Built for non-technical media buyers — making AI-powered decisions feel intuitive.

Design Challenge Solved:

Made ML recommendations transparent for non-technical buyers • Designed confidence-score UI patterns showing *why* the algorithm ranked campaigns high

Outcome: 3x faster campaign workflows across US + UK markets

US+UK
Markets
3x
Faster workflows
My role: Sole product designer. Owned information architecture, interaction design for ML-powered recommendations, confidence-score UI patterns, and campaign builder flows. Collaborated directly with ML engineers on model output formatting and edge-case handling.
Discuss this work
NDA Protected
FinTech · AI-Powered Private Equity

AI Based Private Equity Screener

Designed an intelligent screening platform that uses AI to evaluate and rank potential investment targets for private equity firms. Simplified complex financial data, risk signals, and market indicators into actionable dashboards for deal teams.

Design Challenge Solved:

Built ML explainability into investment decisions • Designed risk-signal visualizations and AI reasoning layers showing deal teams *why* a company scored high

Outcome: 60% faster deal screening with transparent AI reasoning

60%
Faster screening
AI
Explainability layer
My role: Sole product designer. Designed the screening dashboard, risk-signal visualisation, and AI explainability components — showing users why the model ranked a company high or low.
Discuss this work
NDA Protected
Product Strategy · Enterprise SaaS · Organizational Design

Org OS for Flat Teams

Designed a unified platform for managing operations across sales, finance, people, and projects for an organization with zero hierarchy, transparent salaries, and remote friendly model.

Design Challenge Solved:

Designed for transparency without hierarchy • Built workflows for open salary visibility and peer-driven accountability • Created multi-module information architecture enabling autonomy at scale

Outcome: 8 interconnected modules enabling decentralized decision-making

8 modules
Sales, Finance, People, Projects
My role: Sole product designer. Built information architecture across 8 interconnected modules, designed workflows for salary transparency, project allocation, and equity management. Key challenge: Creating interfaces that empower individual autonomy while maintaining organizational visibility — no managers, only clear data and peer accountability.
Discuss this work
NDA Protected
EdTech · PTC University

Redesigning a Learning Platform for 550k+ Global Users

Led UX strategy for a multi-locale education platform across 11 geographies. Integrated accessibility standards and consolidated scattered content types into a cohesive learning experience.

Design Challenge Solved:

Managed complexity at scale across 11 locales • Consolidated scattered content types into unified navigation • Built WCAG AA compliant screen-reader patterns

Outcome: 550k+ global users with accessible, cohesive experience

550k+
Global users
WCAG AA
Compliant
My role: Design lead (team of 2 designers). Ran user research across 4 locales, redesigned navigation and content architecture, implemented screen-reader-compatible patterns, and authored the team's accessibility design guidelines.
Discuss this work
NDA Protected
Telecom · O2 Priority & My O2

Priority Moments & My O2 Self-Service App (Telefonica, UK)

Designed the reward and loyalty experience for O2 Priority, increasing daily engagement for 1M+ subscribers. Separately, rearchitected the My O2 self-service app — streamlining billing, plan management, and support flows.

Design Challenge Solved:

Designed engagement patterns from loyalty data • Redesigned self-service flows to reduce support friction • Built for 1M+ concurrent subscribers

Outcome: 40% fewer support tickets, 1M+ active subscribers

40%
Fewer support tickets
1M+
Subscribers
My role: Product designer within a 4-person team. Owned the rewards feed, offer detail screens, and billing section redesign.
Discuss this work
3×
Faster campaign workflows
40%
Fewer support tickets
60%
Faster deal screening
550k+
Users, WCAG AA compliant

What People Say

Working with Arpit.

"Arpit has a rare ability to take messy, complex ML outputs and turn them into interfaces that non-technical stakeholders can actually use. He thinks in systems, not screens."
Engineering Lead — Sahaj Software
"His accessibility work on our learning platform was exceptional — he didn't just check compliance boxes, he genuinely improved the experience for every user across all our locales."
Product Manager — PTC University
"One of the few designers I've worked with who can sit in a data review meeting, understand model performance metrics, and translate that directly into design decisions the same week."
Data Scientist — Sahaj Software
"Arpit brought clarity to a product that had been confusing users for years. The My O2 redesign significantly reduced our support burden and improved satisfaction scores."
Product Owner — Telefonica UK

Design
Philosophy

How I think about AI products

Understand Before You Design
I start with the data, not the screen. Before opening Figma, I sit with engineers to understand model behaviour, edge cases, and failure modes. The best AI interfaces come from designers who know what the system actually does — not what the pitch deck says it does.
Make the Complex Feel Obvious
A confidence score means nothing if a media buyer can't act on it. I design for the person who doesn't speak "ML" — translating algorithmic outputs into decisions, not data dumps. If the interface needs a manual, the design isn't done.
Show the Why, Not Just the What
Users don't trust black boxes. When an AI recommends something, I design explainability into the interface — showing data provenance, confidence levels, and what the model weighted. Transparency isn't a feature; it's the foundation of trust.
Design the Failure State First
AI is wrong sometimes. I design what happens when a model is uncertain, when data is missing, when confidence is low. Graceful degradation isn't an afterthought — it's the first wireframe I draw. Users forgive errors; they don't forgive confusion.
Ship Rough, Learn Fast
AI products improve through exposure to real users and real data. I push for early releases with tight feedback loops — watching how people actually interact with model outputs, then iterating on both the interface and the model's presentation layer together.
Accessible by Default
An AI dashboard that only works for sighted mouse users is a dashboard that fails half its audience. I build WCAG compliance into the first sprint, not the last — because retrofitting accessibility into data-dense interfaces is ten times harder than designing it in from day one.

My Approach to Data-Intensive Design

1. Make the Invisible Visible

AI models, data flows, and algorithms live in black boxes. My job is to visualize them — showing users the data behind decisions, the logic in the recommendation, the confidence in the prediction. Transparency builds trust.

2. Design for Non-Technical Stakeholders

Complex data shouldn't require a PhD to understand. I design dashboards, screening tools, and recommendation engines for busy people making real decisions under time pressure — not for data scientists in labs.

3. Prioritize Clarity Over Cleverness

Beautiful data visualization is nice. Useful data visualization is essential. I optimize for understanding first, aesthetics second. A clear bar chart beats a confusing interactive animation every time.

4. Test with Real Complexity

Figma mocks don't reveal what breaks under load. I prototype with real data, real edge cases, and real user flows — stress-testing interfaces before they launch to thousands of users.

How I Work

Design for
intelligent products.

I work at the intersection of AI, data, and human experience — designing intelligent systems that are usable, trustworthy, and commercially viable.

Start a Project
AI Product Design
Designing interfaces for ML-powered products — from AI screening tools and recommendation engines to intelligent dashboards that keep humans in control.
UX for Intelligent Systems
Making AI outputs legible — confidence scores, risk signals, explainability layers, and graceful fallbacks when models are uncertain.
Data-Dense Interface Design
Complex dashboards, analytics platforms, and data-heavy enterprise tools — designed for clarity at scale.
Design Strategy
Finding the right problem before solving it. Aligning AI capabilities with real user needs and business outcomes.
Frontend Prototyping
HTML/CSS/JS prototypes that live in browsers — testing AI interactions with real data, not just Figma mocks.
Accessible AI
WCAG-compliant design for AI-powered products — ensuring intelligent interfaces work for screen readers, keyboard navigation, and all abilities.

Tools & technical skills.

Design
FigmaFigJamSketchAdobe XDPrincipleWhimsical
Prototyping & Code
HTML / CSS / JSReact (basic)StorybookCodePenGitHub
AI & Data
Jupyter NotebooksSQL (read)LookerMixpanelConfidence-score UX
Research & Testing
UserTestingMazeHotjarDovetailA/B Testing
Accessibility
axe DevToolsNVDAVoiceOverWCAG 2.1 AAColour Contrast
Collaboration
JiraConfluenceNotionMiroSlackLoom

Remote collaboration.

Based in Indore, India (GMT+5:30). 6+ years of distributed team experience across US, UK, and EU time zones.

Timezone Flexibility
4–5 hours daily overlap with US East Coast, full overlap with UK/EU. Async-first workflows with structured handoffs so nothing blocks.
Documentation-Driven
Detailed Figma annotations, Loom walkthroughs, and decision logs in Confluence/Notion. No context is lost between sessions.
Embedded in Engineering
Daily standups, sprint planning, PR reviews for UI. I work like a teammate, not a contractor handing off static mocks.

Thoughts & Writing

Creative Clarity (Substack)
AI · Machine Learning · Product Design

How I Forbid AI from Hallucinating

Designing safeguards into AI interfaces to prevent hallucination and maintain user trust. A practical approach to building reliability into machine learning products.

Read on SubstackSubstack
Accessibility · Universal Design

How We Made Our Product Web Accessible

PTC University Learning Connector serves users across 11 geographies. A significant proportion rely on screen readers. Here's how we approached inclusive design at scale.

8 min readMedium
Product Strategy

The Design of the Present: Why Shipping Beats Planning

There's a difference between a plan and design. Design for the present means adapting your work to the reality users face today.

5 min readMedium

Let's Scale Together

Full-Time Roles

I'm looking for Head of Design, Principal Designer, or Director roles at AI-focused companies. I bring deep expertise in accessible AI design, system thinking, and team mentorship—ideal for companies scaling their design practice.

  • • Architect accessible, intelligent products from 0 to 1
  • • Scale design systems and lead design teams through growth
  • • Bridge product, engineering, and design to reduce friction
  • • Ensure accessibility is built in, not bolted on

Location: Open to full-time remote, based in India (GMT+5:30)

Let's Talk

Also Open To

Ongoing Partnership

3-6 month ongoing partnerships (20-30 hrs/week) for product scaling, design system evolution, and team mentorship.

Let's discuss →

Advisory & Mentoring

Limited availability for design strategy or mentoring product leaders building AI teams.

Let's discuss →
Email LinkedIn X / Twitter