Nova Learn
Learn Nova is Your Brain's Cheat Code.
The problem Nova Learn solves
The Problem It Solves
Students and teachers face multiple challenges in the learning process:
- Students struggle with consistency and maintaining a disciplined study routine.
- Motivation decreases due to a lack of personalized guidance.
- Most learning tools use a one-size-fits-all approach and do not adapt to individual pace or weaknesses.
- Assessments are manual, slow, and provide delayed feedback.
- Students get distracted easily and lose study focus.
- Teachers lack intelligent tools for automated checking and real-time progress tracking.
How the Solution Helps
Our AI-driven system introduces personalization, automation, and structure into the learning journey.
Hyper-Personalized Roadmaps
- AI generates a custom learning roadmap for every student.
- Continuously adapts to learner speed, strengths, and weaknesses.
Adaptive Study and Assessments
- AI-optimized daily study schedules.
- Automatic revision of weak topics through calendar integration.
- Instant-feedback quizzes and auto-generated assessments.
Behavioral Mastery System
- Ensures a student masters each topic before progressing.
- Reinforcement cycles support long-term retention.
Distraction-Free Study Mode
- Blocks distractions to maintain session focus.
- Converts focused time into XP and rewards.
Full-Loop Integration
- Modules like Courses, Calendar, Quizzes, Notes, and Roadmaps work together seamlessly.
- AI synchronizes performance data across the platform.
- Teachers get automated progress reports.
Makes Learning Easier and More Effective
- Students benefit from an AI tutor, planner, evaluator, and focus assistant in one system.
- Teachers save significant time through automation.
- Parents gain transparency into progress and study habits.
- Learning becomes structured, consistent, and outcome-driven.
Challenges we ran into
Challenges I Ran Into
1. Integrating Multiple AI Features Into One Flow
Connecting roadmaps, quizzes, daily tasks, notes, and behavioral mastery into one AI-driven system was complex.
Each module required different prompt formats and data structures, causing inconsistent outputs.
I solved this by creating a unified AI service layer that standardizes prompts and normalizes all responses before saving them.
2. Stabilizing Gemini API for Structured Outputs
Using the Gemini API to generate roadmaps, tasks, quizzes, and feedback initially produced messy or hallucinated results.
The model often ignored JSON structures.
I fixed this by enforcing strict schema-based prompting, adding cleanup functions, and validating the output before accepting it.
3. JWT Cookie Authentication in Production
JWT cookies worked locally but failed on Render due to cross-domain and secure-cookie policies.
Logins were successful, but authenticated routes returned “Not authorized”.
This was resolved by configuring CORS properly, enabling
SameSite=None
, setting cookies tosecure: true
, and matching HTTPS requirements.4. Building the Real-Time Battle Arena
The battle arena required generating fast quizzes, tracking per-user performance, and computing analytics instantly.
Initial versions lagged and sometimes lost user results.
I optimized this by caching prompts, preloading AI models, and redesigning the scoring logic to make the system both fast and reliable.
Technologies used