Connect And Learn

Connect and Learn is an AI‑powered study copilot that maps what you know, spots weak concepts, and builds smart peer and resource connections to accelerate mastery.

Comment

Inspiration

Modern learning platforms either overwhelm with dashboards or give generic advice. We wanted a transparent, fast, local “study copilot” that helps a student focus.

What it does

  • Maps a student’s courses and weak concepts
  • Predicts short‑term mastery gains (heuristic)
  • Builds an optimized time allocation study plan
  • Enforces sequential deep-focus via a gated checklist
  • Ranks resources by personalized fit
  • Lets you search concepts with natural language
  • Finds peer matches (performance + inferred learning style)
  • Generates a motivational micro-plan via Groq LLM

How we built it

  • React + Tailwind UI
  • Heuristic “ML” modules (predictor, planner, ranker) in ml
  • Similarity engine (similarity.js) for performance & style
  • Greedy time slicing (15‑minute chunks) based on marginal mastery gain
  • LocalStorage-backed study session state with sequential gating
  • Groq API client with model fallback + dynamic discovery
  • Token embedding + cosine similarity for concept search

Challenges we ran into

  • Avoiding duplicate session creation when plans regenerated
  • Similarity score inflation (>10) before clamping
  • Designing AI-like features without misleading users
  • Handling API keys safely in a pure front-end
  • Scope control: many tempting “nice to haves”

Accomplishments that we're proud of

  • Entire personalization layer runs client-side
  • Clear, inspectable formulas instead of vague “AI magic”
  • Robust fallback logic for LLM provider
  • Sequential focus workflow encouraging completion discipline
  • Architecture ready for swapping in real models later

What we learned

  • Simple heuristics + good UX ≈ strong perceived intelligence
  • Early guardrails (clamping, gating, persistence) reduce churn
  • Transparency builds user trust
  • Small composable utilities beat premature abstractions

What's next for Connect and Learn

  • Add Claude / multi-provider suggestion abstraction
  • Real mastery regression (lightweight on-device model)
  • Focus analytics (streaks, variance vs. plan, mastery trajectory)
  • Calendar / Notion export of study plan
  • Multi-student persistence + optional sync backend
  • Rich interactive concept graph (force layout & dependencies)
  • Feedback loop to adapt resource and allocation weighting

Built With

Share this project:

Updates