Twelve years in enterprise software: WCAG-compliant systems for 50,000+ engineers, a QA workflow that saved $3.1M a year, and now the decision layer between AI and the people who have to trust it.
teaching · the method, handed over
colleagues at Automation Anywhere, trained to decide sharper with AI
designed at
four case studies · 2018 to now
01 · Automation Anywhere · Lead Product Designer · 2023–present
Enterprises wanted AI inside their automations and were right to be nervous about it. The decision: don't hide the model behind magic. Make every AI skill something a team can test, monitor, and govern before it runs unattended.
didGovernance-first AI skills: tested, rated, monitored in production
result65%of new enterprise sales · 1.6M+ automations powered
Read the story →
02 · Pegasystems · Senior UX Designer · 2021–2023
QA teams lost a week of every release to manual regression checks. Instead of shipping another tool, I moved testing inside the build workflow they already lived in.
didTesting became a property of the epic, not a separate tool
result400hours back per release · $3.1M+ a year
Read the story →
03 · Siemens · UX Designer · 2018–2020
Vehicle safety testing ran on spreadsheets nobody fully trusted. I redesigned it as one governed data system for 50,000+ engineers, built to WCAG AA.
didOne governed record for vehicle safety data, built to WCAG AA
result$20Minefficiency retired · 50,000+ engineers
Read the story →
04 · The Kabadiwala · End-to-end design · Bhopal, 2019
Scrap pickup in India ran on phone calls and haggling. Before drawing a single screen, I mapped the whole service: rate lists, pickup crews, trust in an unorganized market.
didMapped the whole service; the app came last
result500K+downloads · recognized by the World Economic Forum
Read the story →
each one tests when AI should ask a human
Four things I built to answer questions I couldn't argue my way out of. One is live on the web, one runs inside the company, two began at hackathons.
Reads insurance denial letters, finds the mistake, writes the appeal. Built in 4 days for the Google DeepMind Gemini 3 hackathon. Try the live app.
Upload a messy slide, get a pitch-ready one in about a minute. Built entirely with prompting for the Nano Banana hackathon. Writeup on Kaggle.
Design review agents running inside Automation Anywhere, and the hackathon prototype that beat a roadmap by 18 months.
not a résumé. the actual route.
AI Studio and AI design-review agents. Designing what an agent does alone, and what it brings back to a person.
Moved QA inside the build workflow. 400 hours returned per release; teams trusted tests they could see running.
Designed scrap pickup for Bhopal end to end. 500K+ downloads and a World Economic Forum mention later, it's still running.
Vehicle safety testing data for 50,000+ engineers. The $20M spreadsheet, and the first time I watched trust, not usability, decide adoption.
Freelance and founding-designer work: Thy Marketeers, Medha, RentOnGo. Usually alone, usually everything.
Grew up between two lakes and a lot of small commerce.
the human behind the work
I'm Shivam, a product designer in Bangalore. Mostly on tools that never make it to Dribbble. The through-line is trust: safety engineers trusting a data system, QA teams trusting automated tests, enterprises trusting an agent to act.
Contact
The work I care about most is AI in front of real users, where a wrong decision costs something. Tell me what you're stuck on.