🎯 In Development React Native · Expo EdTech · India

Cracking Govt Exams
Should Feel as
Smart as You Are

I designed and built ExamGuru AI end to end. A mobile prep platform for Indian government job aspirants with a Hinglish AI mentor, gamified quizzes, live current affairs, and a performance engine that tells you exactly what to fix next.

95%
Task Success Rate
4.8/5
AI Mentor Score
13+
Screens Designed
3mo
Research to Build
Welcome
Home
Quiz
Role
Solo Designer + Dev
Platform
React Native / Expo
Duration
3 Months
Domain
EdTech / GovtExam
Targets
IBPS, SSC, Railway
User Research + Information Architecture + Mobile UI Design + Design Systems + Gamification UX + AI Integration + React Native + Figma Prototyping + User Research + Information Architecture + Mobile UI Design + Design Systems + Gamification UX + AI Integration + React Native + Figma Prototyping +
Overview

One app that actually speaks your language

ExamGuru AI is a mobile-first preparation platform for Indian government job aspirants. IBPS, SSC, State PSC, Railway - over 2 crore students prepare for these exams every year, and almost none of the existing tools were built for how they actually study.

The app pairs a Hinglish AI mentor with daily gamified quizzes, live current affairs MCQs, and a performance engine that tells a student exactly where they are losing marks and what to do about it. Designed and built solo from research to a working React Native prototype in 3 months.

95%
Task completion in moderated usability testing
13+
Screens with a single unified token system
4.8
Out of 5 on AI Mentor in user testing
3mo
Research to working React Native prototype
Problem Statement

Generic apps fail India's most motivated students

Every major exam prep app treats all 2 crore aspirants the same. Same syllabus dump. Same English-first interface. Same wall of data with no signal about what to actually work on next.

Students know they need to prepare. They do not know where to start, what to fix, or whether they are actually improving.

Students know they need to prepare. They have no idea where to start.
"Maine teen apps try ki - sab mein sirf syllabus tha. Koi nahi bata raha mujhe ki aage kya karna hai."
Interview participant, 22, IBPS aspirant, Patna
😵
No Direction
Generic syllabus dumps with zero personalization. Students open the app and have no idea where to begin.
🌎
English Only Interfaces
All major apps are English-first. Most aspirants think in Hindi or code-switched Hinglish.
📊
Data Without Insight
Performance screens show accuracy rates but give no actionable signal about what to study next.
⚡
Stale Current Affairs
Current affairs content is buried or weeks out of date, despite being a major exam component.
🎮
No Daily Habit Loop
No streaks, missions, or XP. Users open the app twice and disappear. Retention past day 3 is near zero.
Research

Understanding the aspirant's world

01
🎙

User Interviews

8 semi-structured interviews with aspirants aged 19 to 28 from Bihar, UP, Rajasthan, and Delhi.

  • 7 of 8 use WhatsApp groups as their main study source
  • 6 of 8 struggle to stay consistent daily
  • All 8 cited lack of feedback on weak areas as biggest frustration
02
📱

Competitive Audit

Analysed Testbook, Unacademy, Adda247, and BYJU's Exam Prep across 14 usability and feature criteria.

  • None offer a Hindi-first conversational AI guidance layer
  • Average 4+ taps to reach a practice session
  • Performance dashboards give no actionable next steps
03
📊

Survey Insights

12-question survey to a Telegram group of 2,200 IBPS aspirants. 340 responses in 48 hours.

  • 82% wanted a daily study plan but none had one
  • 71% preferred short 10-minute quizzes over mock tests
  • 68% said an AI mentor beats video lectures
User Personas

Who we are designing for

RG
Rahul Gupta
23 · IBPS PO Aspirant · Lucknow
First AttemptHindi SpeakerLow Confidence
Goal
Crack IBPS PO in his next attempt. Wants structure and a study partner who understands his situation.
Frustration
Feels lost in Quant. Does not know which topics matter most. The apps feel intimidating.
Behaviour
Studies 2 to 3 hours per day, mostly evenings. Responds strongly to visible progress and rewards.
PS
Priya Sharma
26 · SSC CGL Repeat Aspirant · Jaipur
3rd AttemptData DrivenHigh Motivation
Goal
Clear the GK and Reasoning cutoff. Wants to track exactly where she is losing marks each week.
Frustration
Has used three apps and all gave her the same static syllabus. None told her she was getting worse at DI.
Behaviour
Studies 5 plus hours daily. Wants weekly breakdowns with clear trend data. Trusts numbers over advice.
User Journey Map

Mapping Rahul's preparation arc

StageActionThoughtEmotionDesign Opportunity
DiscoverFinds app via a friend's WhatsApp forward"Let me check if this one is actually different"🤔 Curious, skepticalSplash and welcome must signal trust and differentiation immediately.
OnboardPicks exam category and sets a study goal"It is asking me what I want. No app has done this before"😊 Engaged, hopefulGoal setup builds early commitment. Show the AI adapting to inputs in real time.
LearnTakes first quiz and reads a current affairs article"These questions are actually from my exam"🎯 Focused, challengedInstant explanations after wrong answers should feel like lessons, not errors.
ReviewChecks performance screen after first week"Quant is 38%. I really need to fix this"😬 Aware, slightly anxiousAI Insight card must lead with a next action. Coaching, not a report card.
HabitReturns the next day for his daily mission"12 day streak. I cannot break it now"🔥 Motivated, competitiveStreak plus XP is the retention engine. Mission must be completable in one session.
Design Process

From insight to interface

01
Discover
8 user interviews, a 340-response survey, and a competitive audit of 4 major exam apps. Mapped pain points, mental models, and unmet expectations.
02
Define
Synthesized research into 2 personas, a user journey map, and 5 HMW statements. Core principle: guidance first, data second.
03
Ideate
Sketched flows for all core screens. Mapped a flat 4-tab IA to get users to a quiz in under 2 taps from anywhere.
04
Design
Built 13 high-fidelity screens in Figma using a full dark token system. Every component and value tokenized before screen work began.
05
Build and Test
Implemented in React Native with Expo. Ran 5 moderated usability sessions. Iterated on the performance screen and onboarding flow.
DESIGN SPRINT DISCOVER DEFINE IDEATE DESIGN BUILD
Information Architecture

Flat, fast, four tabs

Every core action reachable in 2 taps or less. Navigation locked to 4 tabs before any screen work began.

Home
Good Morning Greeting
AI Insight Card
Today's Mission
Current Affairs Feed
Article + MCQ
Day Streak Counter
Quizzes
Daily Quiz
Topic Practice
Category Pick
Question Screen
Explanation Card
Mock Tests
AI Mentor
Hinglish Chat
Concept Explainer
Formula Cards
MCQ Generator
News Summary
Study Advice
Performance
Weekly Progress Chart
Topic Strength Bars
Speed Per Question
AI Focus Areas
Detailed Study Plan
XP and Level Track
01 Onboarding
/
02 Home Dashboard
/
03 Current Affairs
/
04 Quiz Flow
/
05 AI Mentor
/
06 Performance
Stage 01 / Onboarding

The first impression that actually personalizes

Competitor apps open to a full syllabus with 40 topics and zero guidance. Every interview participant described this as the moment they felt "yeh mere liye nahi hai."

ExamGuru inverts this. Before any content, the user picks their exam, sets a target date, and chooses their study hours. The AI builds their curriculum from these three inputs.

Goal setup builds commitment before habit. Users who set a specific exam date were 3x more likely to return on day 2 in prototype testing.
The onboarding itself signals personalization. Asking "what do you want" before showing any content changes the emotional register of the entire app.
Welcome
Category
Goal Setup
Home Dashboard
Stage 02 / Home Dashboard

One screen. One clear next step.

Students who could not answer "what should I study right now" closed the app within 90 seconds in research sessions. The home screen had to solve this before anything else.

The first thing a user sees is a single AI recommendation. Below it sits a 3-item Today's Mission checklist with XP rewards. Two things to answer: what to fix, and how to start now.

The 12-day streak counter sits above the fold. Loss aversion kicks in once the number is visible. 4 of 5 test users said the streak was why they would return tomorrow.
Previous design had content before insight. Users scrolled past the AI card looking for something to do. Reordering content was the fix, not adding new features.
Stage 03 / Current Affairs

Today's news. Tomorrow's exam questions.

Current affairs is a major component of every government exam and the most underserved feature in every competitor app. Most apps bury it in a tab or show content that is weeks old.

ExamGuru surfaces current affairs on the home screen with a priority ranking system. Each article links directly to an MCQ set. Reading a news story and practicing from it happens in the same flow.

Editorial priority tags (High/Medium) help students decide what to read first based on exam relevance, not recency alone.
The Daily GK Challenge card turns current affairs into a timed 5-minute drill, adding urgency and a completion habit to what is otherwise passive reading.
Current Affairs
Current Affairs Detail
Quiz Question
Quiz Answer
Stage 04 / Quiz Flow

Wrong answers that feel like lessons, not failures.

The standard pattern is a green check or red cross, then move on. Students who got 5 wrong in a row had no way to understand why. The quiz became a test of existing knowledge, not a tool to build it.

Every answer, right or wrong, expands an explanation card with the reasoning behind the correct answer. Correct answers trigger "Nice! Keep going" with XP. Streaks and mastery level are always visible.

Gamified feedback vs plain checkmark: Quiz completion rate was approximately 40% higher with the "Nice! Keep going" micro-feedback in A/B prototype testing.
XP and mastery level always move forward, even when accuracy dips. Beginners stay motivated because progress is always visible.
Stage 05 / AI Mentor

A study partner who speaks your language.

7 of 8 interview participants communicate in Hindi-English code switching. Forcing English-only AI created distance. Users described it as "talking to a professor", not a study partner.

The AI Mentor responds in natural Hinglish. Formal enough for concepts, casual enough to feel approachable. A student can type in Hindi and get back an explanation that sounds like a smart friend, not a textbook.

"Yaar, yeh toh apna lagta hai." Most-repeated sentiment in usability testing. Time in AI Mentor tab was 60% higher in Hinglish mode vs English-only.
Language was not a localization detail. It was a core product decision that changed the emotional register of the entire feature.
AI Mentor Chat
Diagnostic Summary
AI Mentor Insight
Performance Overview
Stage 06 / Performance

Data that coaches, not just reports.

The first version opened with a bar chart and 6 accuracy percentages. Every participant showed visible anxiety. Two closed the screen without scrolling. One said: "Itna sab dekh ke ghabrahat hoti hai."

The redesigned screen opens with a single AI Mentor recommendation: one problem, one specific next action. "Your speed in arithmetic is a bottleneck. We have curated a focused 15-min drill." Charts come after the user already has a plan.

Framing is the design decision. "Quant: 38%" triggers shame. "Your Quant speed is a bottleneck, here is a drill" triggers action. Same data, different outcome.
Data before insight was the original mistake. Reordering the screen, not adding features, was what fixed the anxiety response in testing.
All Screens

Every screen, intentional

13 screens. Every component tokenized before screen work began. Each went through at least 3 design iterations.

Welcome
01 Welcome
Category
02 Category
Goal Setup
03 Goal Setup
Home
04 Home
GK Challenge
05 GK Challenge
Current Affairs
06 Current Affairs
Quiz Question
07 Quiz Question
Quiz Answer
08 Quiz Answer
Goal Setup 2
09 Goal Setup
Diagnostic
10 Diagnostic
AI Insight
11 AI Insight
Performance
12 Performance
Topic Strength
13 Topic Strength
AI Mentor
14 AI Mentor
Design System

Tokens, colours, components

Colour Tokens

bg.primary
#18181b
bg.secondary
#27272a
bg.tertiary
#3f3f46
brand.green
#1a7a3a
brand.red
#fe414d
accent
#c8f542
amber
#f5a623
text.primary
#f4f1ea

Typography Scale

Display H1 to Syne 800
Crack Govt Exam
Heading H2 to Syne 700
Today's Mission
Body to DM Sans 400
Improving this can increase your score by 15%
Label to DM Sans 600 uppercase
AI INSIGHT · DAILY GK CHALLENGE

Components

Buttons
Status Chips
AI INSIGHT+15 XPHIGH PRIORITY
Progress Bars
Quant38%
Reasoning88%

Spacing and Radius

4px Base Scale
4
8
12
16
24
Border Radius
4px
8px
12px
Full
Outcomes

What we achieved

95%
Task completion across 5 moderated usability sessions
4.8/5
Average satisfaction score on AI Mentor, highest-rated feature in testing
+60%
Time in AI Mentor tab with Hinglish mode vs English-only in A/B prototype testing
+40%
Quiz completion rate with gamified feedback versus plain correct or wrong indicators
2min
Average time to first quiz, versus 8 minutes average across 4 competitor apps
5/5
Test users confirmed they would return next day after completing their first daily mission
Reflections

What this project taught me

01
Language is a design material
Choosing Hinglish was not a localization task. It was a core product decision that changed the emotional register of the entire app. Every string of text shapes whether a user feels seen or alienated.
02
AI UX is not the same as chat UX
A well-designed AI feature feels invisible. The AI Mentor works because it appears in context, speaks the user's language, and leads with an action rather than an explanation.
03
Gamification requires emotional stakes
Streaks and XP only work when the user cares about losing them. The streak counter works because it is the first thing visible. Loss aversion is strongest when the potential loss is front and centre.
04
Building solo sharpens decisions
Being both designer and developer forced brutal prioritization. Every animation, every component, every token had to justify its implementation cost. Constraints produced better decisions.