Products
Intelligence
Transform thousands of student conversations into structured product intelligence.
Internal Research Infrastructure • LearnzyLabs
Product Link
| Status | 🟢 Active |
| Category | Internal Research Infrastructure |
| Built By | Himanshu Gupta |
| Role | Solo Builder |
| Stack | Python · React · TypeScript · Supabase |
| Purpose | Structured Product Intelligence |
Overview
Intelligence is an internal research platform developed inside LearnzyLabs to support product decisions for Learnzy.
Rather than relying solely on interviews or intuition, the system continuously analyzes large collections of public student discussions and converts them into structured behavioral signals.
It serves as the research layer behind product development, content strategy, and user understanding.
Instead of asking “What do I think students need?” the goal is to answer “What are students actually discussing?”
Why I Built It
While building Learnzy I realized that every product decision eventually came back to the same question:
What is the student actually struggling with?
Individual conversations were useful, but they rarely represented the larger picture.
Thousands of discussions across Reddit contained the answers, but reading them manually wasn’t scalable.
I built Intelligence to continuously organize those conversations into something searchable, measurable and reusable.
The Problem
Student conversations are unstructured.
The same problem appears hundreds of different ways.
Some students ask directly. Others describe symptoms. Others simply vent.
Without structure, it’s almost impossible to understand what problems consistently emerge across thousands of discussions.
The Solution
Intelligence automatically organizes public student discussions into structured behavioral categories.
The platform allows our team to:
- Explore discussions by behavioral category
- Track trends over time
- Search across large datasets
- Export filtered datasets for AI workflows
- Build content from actual student language
- Support product decisions with evidence instead of assumptions
Features
Data Collection
- Multi-community Reddit ingestion
- Time-based filtering
- Search
- Subject filtering
Behavioral Classification
- Anxiety & Panic
- Doubt & Advice
- Revision & Notes
- Mock Tests
- Schedule & Routine
- Coaching & Batches
- Silly Mistakes
- Controversy & Leak
Research Dashboard
- Behavioral distribution
- Timeline visualization
- Submission explorer
- Keyword analysis
- Trend monitoring
Data Export
- CSV export
- LLM context export
- Structured datasets for downstream AI workflows
How It Works
Student Communities
↓
Python Collection Pipeline
↓
Cleaning & Classification
↓
Behavior Categories
↓
Research Dashboard
↓
CSV / AI Export
Design Philosophy
Intelligence isn’t designed to generate reports.
It’s designed to support research.
Every screen answers one question:
Can this help us understand students better?
The interface intentionally stays simple because it is used repeatedly during product development and content planning.
Rather than maximizing visual complexity, the focus is reducing friction for internal research.
Technical Challenges
The hardest problem wasn’t collecting Reddit posts.
It was designing a classification system that consistently groups thousands of discussions into meaningful behavioral categories while keeping the system flexible enough to evolve as new topics emerge.
Another challenge was balancing analytical depth with usability.
The dashboard needed to remain powerful for research while still being fast enough to use every day during product discussions.
How We Use It
Intelligence is used throughout the Learnzy workflow.
Product
- Identify recurring student problems
- Prioritize new features
- Validate assumptions
Research
- Discover behavioral patterns
- Measure changing concerns
- Build internal datasets
Content
- Find recurring language students naturally use
- Generate video ideas
- Validate messaging
Privacy
The repository intentionally excludes all collected datasets.
Only the application, processing pipeline, and schemas are included.
The underlying research datasets remain private because they are used internally for product research.
What I Learned
Building Intelligence fundamentally changed how I think about user research.
Individual conversations create empathy.
Large collections of conversations reveal behavior.
Instead of looking for isolated opinions, I now look for recurring patterns that appear consistently across thousands of people.
That shift has influenced every product decision I make.
Future Direction
Future work includes:
- Semantic search
- Behavioral clustering
- Trend prediction
- Automatic hypothesis generation
- Cross-platform intelligence
- AI-assisted research summaries