Trajectory

The problem

Free learning material is everywhere: university lectures, YouTube, practice problems. What most students don’t have is a path through it. If you want to learn calculus, investing, or programming, the hard part is knowing what to study first, what to study next, and whether you actually understand it.

What we built

Trajectory starts from a goal (math, finance, or coding) and your current level, and builds a plan around it.

• A content library. We scraped and organized lectures, articles, videos, and problem sets for each subject, including MIT OpenCourseWare and Yale lectures, Art of Problem Solving problems, and Investopedia articles.
• An AI pipeline that enriches each lecture, builds a graph of the concepts and how they depend on each other, generates a study plan from your goal and level, and writes practice questions.
• A dashboard with your current plan, next action, accuracy over time, concept mastery, and a draggable knowledge graph.
• Study sessions with a timer. Afterward you rate the session and say what worked, and the AI rewrites the weeks you haven’t done yet. If you liked videos, for example, upcoming weeks lean on videos. Completed weeks stay as they were.

My role

I built Trajectory with Mrinal, Arnav, Alex, and Aryan. My parts:

• The finance content: scrapers for Investopedia articles, finance education sites, and YouTube videos.
• Study sessions: the session timer, the engagement rating at the end, and saving sessions to the database.
• Plan selection and adjustment: choosing a curriculum on the subjects page, the logic that adjusts future weeks after a session, and a fix for duplicate plans.
• AI explainer videos: a client that sends a script to Opennote’s video API and waits for the finished video.

I also helped get the app deployed.

Outcome

Trajectory is live as a demo. The code is in a private repository.