Backtrack

Lost in class? Backtrack shows you why you’re stuck and gives you the quickest steps and resources to understand the topic again.

Comment

Backtrack

Inspiration

Students often fall behind in a class not because they are lazy or because they need more content, but because they are missing earlier building blocks. A student may be confused by a current topic, but the real issue is that they forgot or never fully understood the prerequisite concepts behind it.

Most AI tutoring tools explain the topic directly. Backtrack takes a different approach: it helps students figure out what they need to review first before trying to understand the current lesson.

Our main idea is simple:

Students are not always behind. Sometimes, they are just missing one step.

What It Does

Backtrack is a prerequisite-aware learning recovery web app. It diagnoses why a student is stuck and builds a personalized path to help them catch up.

Students can upload course materials such as a syllabus or lecture slides. Backtrack then extracts the main course topics, identifies likely missing prerequisite concepts, explains why those prerequisites matter, recommends free learning resources, generates a short readiness quiz, and provides a focused tutor chat.

Backtrack is not a generic AI tutor. It is a course-aware learning recovery system.

Core Features

  • Upload syllabus or lecture slides
  • Extract course topics from uploaded materials
  • Identify missing prerequisite concepts
  • Explain why each prerequisite matters
  • Recommend 2–3 free learning resources
  • Generate a short readiness quiz
  • Provide a focused tutor chat for follow-up questions

The Problem

When students are confused, they often do not know what they need to review first. They may keep watching videos, reading notes, or asking AI to explain the same topic again, but the explanation still does not make sense because the missing knowledge is earlier in the learning chain.

The real problem is often not:

“I need more content.”

It is:

“I am missing the prerequisite knowledge that makes this topic understandable.”

Backtrack solves this by helping students work backward from the topic they are struggling with to the foundation they need to rebuild.

Why This Idea Matters

Most learning tools focus on answering questions or generating explanations. That is helpful, but it does not fully solve the recovery workflow for students who are already behind.

Backtrack focuses on this missing workflow:

  1. Identify the target topic the student is struggling with
  2. Diagnose likely missing prerequisite concepts
  3. Explain why those concepts matter
  4. Recommend high-quality free resources
  5. Check readiness with a short quiz
  6. Offer follow-up tutoring only after the gap is identified

This makes the learning process more focused and less overwhelming.

How We Built It

We built Backtrack as a web application using a modern full-stack architecture.

Tech Stack

  • Frontend: Next.js 14+ with TypeScript
  • Backend: FastAPI
  • AI: OpenAI-compatible model endpoint
  • APIs: YouTube Data API for video resource search
  • Document Processing: PDF parsing for syllabus and lecture slide uploads

System Components

  • Document Ingestion: Handles uploaded syllabus and lecture files
  • Topic Extraction Engine: Extracts important course topics from the uploaded material
  • Prerequisite Inference Engine: Identifies likely missing prerequisite concepts
  • Resource Recommendation Engine: Finds free learning resources for each prerequisite
  • Quiz Generator: Creates short readiness checks
  • Tutor Service: Provides focused explanations and follow-up help

The backend handles AI orchestration, PDF parsing, topic extraction, prerequisite inference, quiz generation, and tutor endpoints. The frontend provides the student-facing experience from upload to recovery plan.

Challenges We Ran Into

One major challenge was document parsing. Lecture slides and PDFs do not always extract cleanly, so we had to design around imperfect input.

Another challenge was making the AI output useful instead of generic. If the prompts were too open-ended, the results could feel vague or incorrect. We had to constrain the prompts and focus the app around strong demo cases.

Resource recommendations were also difficult because YouTube search results can be noisy. We needed to think carefully about how to return resources that are actually useful for learning the prerequisite, not just loosely related to the topic.

Finally, connecting the frontend and backend under hackathon time pressure was a challenge. We had to keep the scope focused so the core flow worked reliably.

Main Challenges

  • PDF parsing reliability
  • LLM prerequisite accuracy
  • Avoiding generic AI responses
  • YouTube search noise
  • Frontend and backend integration under time pressure

Accomplishments We’re Proud Of

We are proud that Backtrack creates a clear “gap reveal” moment for the student. Instead of simply explaining the current topic again, the app shows the missing foundation that may be causing the confusion.

We are also proud that the user can go from uploading a course file to receiving a focused recovery plan in a short amount of time.

Our strongest accomplishment is the project’s narrative and workflow: Backtrack helps students stop guessing what to study and gives them a clearer path to catch up.

Key Accomplishments

  • Built a working upload-to-recovery-plan flow
  • Created a prerequisite-aware learning experience
  • Generated personalized recovery plans from course materials
  • Added quizzes to check readiness
  • Designed a focused tutor chat instead of a generic chatbot

What We Learned

We learned that prerequisite inference is powerful, but it also needs guardrails. AI can sometimes make imperfect assumptions, so the product should acknowledge uncertainty instead of pretending every answer is perfect.

We also learned that a strong demo depends on good constraints. By using curated demo files, focused prompts, and clear test cases, we were able to make the experience more stable and understandable.

Most importantly, we learned that educational AI tools should not only answer questions. They should help students understand where their confusion is coming from.

What’s Next

In the future, we want to make Backtrack more visual, more personalized, and easier to share.

Planned improvements include:

  • Readiness score or mastery indicator
  • Concept dependency graph visualization
  • Ability for students to mark prerequisites they already know
  • Exportable recovery plan as a PDF or shareable link
  • LMS integrations
  • Instructor dashboard for identifying common class-wide learning gaps

Demo Story

A student misses class or struggles with a current topic. Instead of asking for another generic explanation, they upload their syllabus or lecture slides into Backtrack.

Backtrack identifies the topic they are stuck on, works backward to find the missing foundation, explains why that prerequisite matters, recommends free resources, and gives the student a short quiz to check if they are ready to move forward.

The goal is not to overwhelm the student with more content. The goal is to give them the shortest path to catch up.

Built With

Share this project:

Updates