LearNexus
Build differently to learn clearly
Created on 12th April 2026
•
LearNexus
Build differently to learn clearly
The problem LearNexus solves
Traditional online education is notoriously static—a "one-size-fits-all" model. If a student misses a day, gets sick, or struggles to grasp a concept, the course syllabus marches on without them, often leading to drop-offs. Lectures lack personalization and fail to adapt to an individual's learning style or optimal pace.
LearNexus solves this by introducing a deeply personalized, highly-resilient Agentic AI Tutor. Instead of static videos, our AI dynamically builds a custom curriculum and syncs it to your Google Calendar.
Fell behind? LearNexus detects a low quiz score and automatically injects custom revision lectures.
Missed a day? The schedule recalculates and shifts future lessons forward without skipping any content.
Confused mid-lecture? The real-time Doubt Resolver answers queries strictly using the immediate lecture context to prevent AI hallucinations.
Essentially, LearNexus democratizes the luxury of a 24/7, hyper-adaptive private tutor.
Challenges we ran into
-
LLM Rate Limits & Orchestration: Running 8 different highly-specialized AI agents (via CrewAI) simultaneously led to frequent rate limits and timeouts. I solved this by implementing a custom multi-provider LLM fallback strategy (using LiteLLM) to seamlessly reroute traffic to backup models if a primary API endpoint failed.
-
Session Resilience & State Sync: Keeping the AI backend cleanly synced with the user's progress across browser restarts was tricky. I tackled this by overhauling the memory layer, combining precise Pydantic state management with real-time WebSocket events and persisting them to Supabase.
-
Audio Pipeline Timeouts: During deployment, generating and uploading audio files to Cloudinary caused serverless function timeouts. I bypassed this bottleneck entirely by migrating to a fast, free, browser-native Web Speech API, which reduced backend heavy-lifting and cut latency down to almost zero.
-
UI/UX Accessibility Pivot: The initial prototype leaned heavily into a dark "glassmorphism" aesthetic that suffered from contrast and accessibility issues. I executed a rapid UI pivot, redesigning the frontend into a bright, gamified, and WCAG-compliant dashboard using Tailwind and Framer Motion—keeping the "wow" factor while making the app usable for everyone.