GyanGram Turning UPSC preparation from scattered study into a guided mastery journey.
A Flutter-based UPSC prep app where every card, article, and PYQ feeds a live mastery map of your exam preparation. Built entirely solo: design, code, content, and systems, using AI-augmented workflows.
One thought while scrolling a slide deck.
In early 2025, I was scrolling through a NotebookLM slide deck. A thought struck: millions of UPSC aspirants scroll through Instagram and Reels every day — what if that exact scroll behaviour led to learning instead of entertainment?
That question, shared with Claude in a conversation, became GyanGram. The idea was deceptively simple: convert the behaviour, don't fight it. But building the product revealed the real problem — and it wasn't what the original idea assumed.
"UPSC aspirants don't fail because they lack content. They fail because they don't know what they actually know."
— The insight that changed the product from a content feed into a learning system
India has infinite UPSC content. That is not the problem.
YouTube channels, PDFs, printed notes, coaching videos — UPSC aspirants are surrounded by content. They still fail. Because content access was never the gap.
No live picture of mastery
Aspirants read topics but have no way to know whether they've actually learned them or just seen them. Passive reading feels productive. It isn't.
Social media wins the same screen time
A 30-minute free window ends up on Instagram, not study. The phone is the problem and the solution. GyanGram bets on using the behaviour, not fighting it.
Revision is broken
Most students never return to what they studied. Notes are written and forgotten. Wrong PYQs are accepted and repeated. There's no system driving them back to weak areas.
Heavy formats for a short-attention era
3-hour video lectures and 200-page PDFs are designed for a different generation of learners. Most aspirants today study in fragments — apps and commutes and lunch breaks.
The honest version — how Version 1 became what GyanGram is now.
The first version had three tabs: Home, Explore, Profile. Explore was a scroll feed. The app existed. But it had no reason for users to return.
"Why would a user keep coming back to this app?"
— The question that turned a content feed into a learning system
The answer: only if the app showed them their progress and told them what to do next. That question led to the Learn tab as a syllabus mastery map and the Home tab as a daily coach — not a generic dashboard.
Three more pivots followed, each driven by a real problem hit during building:
The image pivot
Early cards were AI-generated images. Beautiful — but fragile. One wrong fact required a completely new image. No searchability as content scaled. I researched Flutter's widget rendering and built ~50 reusable card templates. Now a JSON file creates an entire card deck in seconds. AI images remain for creative visuals where precision matters less than impact.
The depth discovery
Cards teach facts. Articles teach understanding. I built an interactive markdown reader and linked articles to cards via long press — the aspirant can go deep on any concept, then return to the card stream. Two-layer learning: quick recall plus deep comprehension. That pairing is what UPSC Mains actually requires.
The mastery architecture
I analysed 15 years of UPSC Prelims PYQs and structured the syllabus from that signal. Every topic became a node — linked to cards, articles, quizzes, and PYQs. Progress is tracked at node level. The result: the syllabus becomes a live mastery map, not a static reading list.
Four tabs. One learning loop.
Each tab maps to a distinct phase of learning. Nothing overlaps. Every interaction in any tab feeds back into the mastery system.
Study → Recall → Practice PYQs → Diagnose gaps → Revise → Earn mastery.
— The loop every tab is designed around
Home
Your daily coach. Not a generic dashboard — a prioritised list of exactly what to do today.
- Goal-based daily study plan
- Spaced repetition scheduling
- Previous-day performance signals
- Weak areas and due revisions surfaced
Scroll
Discovery with purpose. 108+ cards — scroll to learn, not to waste time.
- MCQ, flashcard, true/false, fill-in, match templates
- Horizontal decks for concept depth
- Long press → linked article handoff
- Like, bookmark, share per card
Learn
The syllabus as a mastery map. Structured study from subject down to micro-topic.
- PYQ-signal-based syllabus structure
- Mastery node per topic
- Linked articles, cards, quizzes, PYQs
- Completion and progress tracking
Review
Active repair, not a static resource shelf. Return to what needs attention.
- Due topics and weak areas
- Wrong PYQs queue for retry
- Bookmarked cards and articles
- PYQ trend analysis and exam tools
What I chose not to build — and why that matters more than what I did.
The easiest product mistake is adding features. Every item below was considered, prototyped mentally, and deliberately rejected.
All AI-generated image cards
Visually rich but fragile at scale. One factual error requires a completely new image. No searchability as the card count grows. Switched to Flutter widget templates — accuracy, speed, and searchability together.
Heavy gamification
Streaks, leaderboards, and reward systems optimise for app-opens, not learning. UPSC Mains is a six-hour writing examination. Streaks kept for psychological value. Everything else removed.
Generic MCQ banks
Generic MCQs give false confidence. PYQs are the actual exam signal — 15 years of them analysed to structure the syllabus. Interactive card formats (fill-in, match, true/false) do what MCQs do without abandoning depth.
Social comments on cards
GyanGram is a study tool, not a social platform. Every aspirant's learning journey is different. Comments would fragment focus without adding learning value. Deliberately not built.
Text-based card templates over images
Searchability at scale. As cards reach thousands, students need to find specific topics instantly. Image cards are unsearchable. Flutter widget templates give full text search from day one.
JSON import for card decks
Speed of content creation without sacrificing structure. One structured JSON file generates an entire horizontal card deck in seconds. Reviewed and edited as needed. Content velocity without quality compromise.
Flutter for all platforms from day one
Android, iOS, web, and Windows from a single codebase. UPSC aspirants span urban smartphone users to desktop-first coaching students. No user excluded by platform choice from the beginning.
Launch after content seeding, not before
A content-first release means every early user gets a complete experience. Launching with skeleton content trains users to expect an incomplete product. Quality before reach.
One person. Design, tech, content, and systems — all of it.
GyanGram is not a side project. It is a complete product operation: a design system, a content pipeline, an admin CMS, and a multi-platform application — built without writing traditional code, using AI-augmented development throughout.
The key philosophical constraint: AI as infrastructure, not as author. AI generates card visuals, assists with code, and speeds up structuring. But every factual claim, every PYQ explanation, every article is authored with six years of domain knowledge. Trust is the product.
Building a product solo forces every decision to be explicit — there is no PM to escalate to, no designer to delegate the hard trade-offs to. Everything in this case study was decided, owned, and executed by one person. The founder's six years of UPSC preparation (Prelims cleared 3 times across 5 attempts) is not a background detail — it is the user research that shaped every feature decision, every scope exclusion, and every content philosophy choice. That is a different kind of product experience than roadmap collaboration. It is closer to founding PM work.
Sequenced by user value, not feature count.
-
NowIn progress
Content seeding
Building the initial content base across Geography, Polity, History, Economy, and Environment before public launch. Quality before velocity.
-
Month 1Next
Play Store launch (Android)
Public release when content demonstrates the full scroll → learn → review loop across at least three subjects.
-
Month 3–6Planned
Feedback-driven iteration
Roadmap shaped by real user behaviour. Target: 10,000 active users before adding any major new feature.
-
Year 1Planned
1 lakh users · iOS · Hindi vernacular
Expansion to iOS and web. Hindi vernacular card support — content creators needed, Flutter feature is one build away.
-
LaterPlanned
AI learning companion
AI integrated into the study journey — helping users understand their gaps and what to do next. Complement to content, not a replacement for it.
Interested in GyanGram — as a user, collaborator, or as a PM case study?