← All work Case 04 / 07 — EduVision
04 / 07
Case study 04 — EdTech · consumer · gamification

EduVision

A career-prep platform for Indian students who are 92 days from placement and already frightened. The whole design problem was the fear, not the features.

Outcome Five tools in one desktop-first platform — Resume Builder, Test Prep, AI Interview, Skill Tracker and Expert Network — an XP economy with a 12-item store, and a 1,100-line written system spec that includes a vocabulary the interface is forbidden from using.
Product
EduVision — LearningVision suite
Users
Indian college students, 18–24
Role
Sole designer, system + all screens
Deliverable
System spec v1.0 + built app
The problem

Every other product in this portfolio is used by a professional doing their job. This one is used by a final-year student who is scared. Indian placement season is intense, unstructured and public — everyone knows who got an offer and who didn't — and the student has no idea what to do next.

That reframes the brief entirely. The ask was “combine five tools into one platform”. The actual job was to make an anxious person feel they are making progress today, without lying to them about where they stand.

Which sets up the hard part. A dense BI dashboard is the honest way to show someone their real position. It is also the fastest way to overwhelm them.

Index card — who this is for
RS Rahul SharmaFinal year B.Tech CS · VIT VelloreTarget: Software Engineer
Days to placement92
XP · level2,400 · Hustler (Lv.3)
Streak7 days
Resume68% complete
Leaderboard3rd of 3 shown
Weakest subjectAptitude · 52%

Every one of those last three numbers is a chance to make him feel worse. That is the design problem.

The decision that made it work

A list of words the product may not use

Tone is usually the last thing anyone specifies and the first thing that drifts. So the system spec includes a banned vocabulary and a set of mandatory reframings, written before any screen — because for this user, the language is the interface. A student who reads “Test Failed” closes the tab and does not come back.

Never appears in the UI banned
Error Failed Invalid Unfortunately
Mandatory reframings
Test failedKeep practicing!
ContinueBuild Resume
AI InterviewPractice Mode
Strengths before weaknesses, always

A structural rule, not a copy rule: on every results screen the strong subjects render above the weak ones. Same data, same honesty, opposite emotional outcome — and it costs nothing to enforce.

One primary action per screen

Exactly one #BD1313 button anywhere on a screen. A paralysed user does not need options, they need a next step — and the rule is what stops a dense dashboard turning into a wall of equal choices.

Named, not neutral

The interviewer is “Aria”, and starting a session says “take a breath”. Giving the AI a name and a manner is what lets it be practice rather than assessment.

The dashboard

Duolingo warmth, BI density

A 64px icon-only sidebar and a three-column grid, specified to fit one viewport — everything needed to answer “what do I do today”. In the build that rule had to bend; see below.

eduvision.aivision21.com/dashboard — live, interactive
The dip is not hidden

The score trend shows the real sequence — 62, 58, 71, 68, 74, 79, 74 — including the drop at test two, ringed rather than smoothed. Hiding a bad week would make the good ones meaningless.

Subjects run best to worst

English 88 first, Aptitude 52 last. The weak subject is still there in red — it just isn't the first thing he reads about himself.

Locked badges stay visible

Three of eight unlocked, and the five he hasn't earned are shown greyed with their names — a target rather than a hidden surprise. The tasks card does the same job daily: three ticked, one small thing left, +30 XP.

The part I argued about

Gamifying a frightened person

XP, streaks and a leaderboard are motivation mechanics borrowed from products people use for fun. Pointed at someone terrified of not getting placed, the same mechanics become pressure — a broken streak is one more failure, and a leaderboard is a daily reminder that two people are beating you.

I kept them, with three constraints: XP is only ever awarded, never deducted for inactivity; the ladder is named in aspiration rather than rank (Starter → Champion, no numeric position); and XP buys real relief in a store — a Streak Shield, an AI review, a resume template — so effort converts into something that reduces the workload rather than just a bigger number.

Starter
0–499
Rising
500–1,499
Hustler
1,500–2,999
Rahul is here
Elite
3,000–4,999
Champion
5,000+
What earns XP
Daily login+10
Complete any task · add a skill+20
Save a resume section+30
Take any test+50
Practice interview session+75
Score 75%+ · 7-day streak+100 / +200

Taking a test earns XP whatever the score. Effort is what's rewarded; the score bonus sits on top.

The store is where the design argument lands

Twelve items across six categories, and the most-bought are deliberately the ones that lower pressure rather than raise status: a Streak Shield at 500 XP forgives a missed day, an AI Resume Review at 400 XP does work for you, an XP Boost Rocket at 600 XP accelerates the ladder.

Profile frames and Elite badges exist too, priced similarly — but a student who is behind can spend their way to relief instead of decoration. That is the difference between a game and a support tool wearing a game's clothes.

Streak Shield · 500 AI Review · 400 Boost Rocket · 600
Decisions worth defending

Five calls, and what each cost

DecisionReasoning · trade-off
“One viewport, never a page scroll” — a rule I lostThe spec forbade page scrolling outright: a scrollbar on a dense dashboard implies more bad news below. It held everywhere except here. Eleven widgets would not fit 900px honestly, so the dashboard shipped with overflowY: auto and the rule became inner-scroll-first rather than absolute. I'd rather show the rule breaking than pretend the constraint survived contact with the content.
Emoji in every section headerDeliberately un-corporate, for an audience that finds corporate intimidating. Costs credibility with anyone senior looking over the student's shoulder — and it would be wrong for every other product in this portfolio.
Leaderboard shows three rows, user always in themA student ranked 340th learns nothing from seeing 340. Costs the honest global picture — arguably the weakest compromise on this page, and the one I'd test first.
The AI interviewer has a name“Aria” with a green live dot, opening with “take a breath”. Framing it as a person practising with you, not a system scoring you. Costs the sense of a real assessment, which is exactly the point.
A 92-day countdown on the hero cardThe scariest number in the product, shown on purpose — because a student who doesn't know their timeline can't plan. Paired with a streak pill so the two read together: time is short, you are moving.
What I'd change

I designed an anxiety-aware product without testing it on an anxious student. Every rule here — strengths first, banned words, XP that only goes up — is reasoned rather than observed, and the whole thing rests on an assumption about how a frightened 21-year-old reads a dashboard. Five students and a think-aloud session would have been worth more than half the spec.

It is also desktop-first for a generation that lives on a phone. That was the brief, and I should have pushed on it.

What it proved

Emotional state is a design constraint you can write down. Once “the user is frightened” is in the spec next to the colour tokens, it produces testable rules — an ordering rule, a vocabulary, a limit of one primary action — rather than a vague instruction to be friendly.

It is also the clearest evidence in this portfolio that I can design for a consumer audience, not only for domain experts doing their job.

Next case study — 05 of 07
PeopleVision

Ten role-gated workforce reports across 847 employees.

Read it →
← Previous: Learning Vision
All work About Email me