← Back to all work

Case study 02 · Product design · Mobile app

Faster help, trusted information. A safety app for India's everyday emergencies.

Role
Sole UX Designer
Company
Tejis.ai
Platform
Android
Timeline
3 weeks to first handoff
Team
AI team, mobile devs, stakeholders
Status
Production-ready, 8 test builds

In 30 seconds

The product and the people it serves

When something dangerous happens nearby in India, most people call someone. They post on WhatsApp. They hope someone sees it. Citizen App was built to change that: a hyperlocal safety platform where people report incidents in real time, trigger SOS alerts and stay informed with verified local news.

The users are everyday citizens, commuters and families in Tier-1 and Tier-2 cities. People who need safety information fast, not another news app.

The problem, split three ways:

6modules designed
8APK test builds
3weeks to handoff
1designer, me

My role, and how the team worked

I owned UX end to end for all six modules, from problem to developer handoff. Requirements and initial research came from the Head of AI. The AI and ML team built incident clustering and verification, the mobile developers shipped the Android builds, and the stakeholders set the business direction. My job was to turn all of that into a clear, usable, trustworthy experience.

The timeline was three weeks to first handoff, so the process had to be lean: understand the PRD and the Indian safety context, map the flows with stakeholders, design in quick AI-assisted prototypes first, then refine the final screens in Figma through team feedback, developer feasibility checks and eight rounds of APK testing.

Three insights that shaped everything

Module 1: reporting an incident

The hard part: the person filling this form is stressed, scared or in danger. My first version put everything on one long screen. The team flagged it immediately, and they were right. To a person who just witnessed an accident, a long form is not a tool. It is an obstacle.

So I broke it into steps, one question per screen. What happened? How urgent is it? Add details. Review. Submit. Each screen asks one thing a stressed person can actually answer.

The first version of the incident report: a single long screen with every field at once The final incident report, with plain location names and a clearer layout
The incident report: the single-screen first version, and the final flow.

Module 2: a feed you can trust

The feed mixes citizen reports with verified news. My first instinct was to separate them into tabs, because they carry different levels of trust. I designed it that way twice. Then the investor pushed back and asked for one mixed feed. That was a real business constraint, and the direction changed.

But the trust problem did not go away just because the tabs did. So I moved the trust signal onto every card instead. Each post carries a badge showing exactly where it stands: under review, verified, approved or not approved. Once the AI confirms a report, the badge updates on its own. Users always know what they are reading.

When I could not control the structure, I controlled the signals.

Honestly, the badge system solved the problem better than the tabs would have. It was appreciated by stakeholders, and it kept the feed dense without making it misleading.

Feed iteration 1 Feed iteration 2 Feed iteration 3 Feed iteration 4, with per-card verification badges
The feed across four versions, moving the trust signal onto every card.

Module 3: SOS, the highest-stakes flow in the app

A person triggering SOS may be panicking, injured or in danger. Every extra tap and every unclear label has a cost. The first version had a plain SOS button. One tap, alert sent. The gap: police, ambulance and fire services respond very differently, and a generic alert gives responders nothing to act on. But adding an extra screen to pick the type felt wrong too.

The answer was to put the emergency type inside the gesture itself. Swipe up for a medical emergency. Swipe down for anything else. Zero extra screens, zero extra taps, and the responder still gets the context they need.

The SOS trigger: swipe up for a medical emergency, swipe down for anything else The SOS countdown, with contact avatars appearing as each person is notified
The swipe-based SOS trigger, and the countdown that shows real people being reached.

Module 4: the map, and the feature I proposed myself

A safety map has to answer two different questions. What is happening right now? And which areas are dangerous over time? One dropdown switches between the two: an incident view with markers for live events, and a hotspot layer that turns the map into a heat map of the last three months. Tapping a hotspot shows an AI summary in plain language: "45 incidents in 3 months: 20 accidents, 15 fires, 10 floods." Insight, not just data.

The bigger idea came from the third insight, that safety in India is a family concern. I proposed putting family members on the same map as the incidents, so you see where your people are and what is happening around them in one view. That proposal reframed the app from a personal safety tool into a family safety platform.

The trade-off: the Family Safety Network was fully designed and prototyped, then cut from the first release for time. It remains ready for a future version.

The incident view: live event markers on the map The hotspot layer: a heat map of the last three months of incidents
The family drawer: family members shown on the same map as incidents
One map, three answers: live incidents, a hotspot heat map, and the family drawer.

Module 5: an AI legal advisor for people who never had one

Most people in India do not know their basic rights. What can a police officer ask of you? What documents must you carry? The feature had to make legal help feel possible for someone who has never spoken to a lawyer and would not know how to start.

An empty chat box is intimidating, so the screen opens with topic chips: Traffic Rules, Police and Arrest Rights, Women's Rights, Cyber Fraud, Workplace Rights, Consumer Rights. Tap one and the conversation starts. The voice mode shows an example before you speak: "Say something like: What are my rights if police stop me?" That one line teaches people how to ask.

Module 6: a profile that closes the loop

Most profile pages are settings pages in disguise. This one tracks outcomes. Every report a user submits goes through verification, and My Reports shows exactly where each one stands, with a notification badge when something changes. Badges show what you did and when, and unearned ones show a progress bar toward unlocking them. Contribution you can see, not just a points total.

The result

How I would measure success

The app never reached the public, so these numbers were never collected. This is what I defined as success for each high-stakes flow:

Reflections, and what would come next

The honest lesson: a good design decision, backed by data and logic, can still be overruled by seniority. The feed tabs were the right call and they still got cut. The skill is not only making the right decision. It is finding a way to serve users inside the constraints you are given. The badge system was my answer to that.

The SOS module taught me how to design for panic: not what users need, but what they are capable of doing under extreme stress. I will carry that into every project.

What I would do differently: real field testing, not just internal feedback. People actually using the app on a busy street. And more time on edge cases, empty states, error flows and offline behaviour, because that is where a design proves it is truly done.

Given more time, the roadmap was clear:

Next case study Nucleus Design System →