IdeaStruct

The problem

The value in a pile of notes isn't the words — it's the connections between them, and those stay invisible in flat text. IdeaStruct makes that structure explicit: describe what you know in plain language and get back a living graph you can see and query.

What I built

• Built a Flask web app that turns natural language into dynamic knowledge graphs — you write, it maps.
• Used OpenAI GPT for the NLP layer: extracting entities and the relationships between them from raw text.
• Backed it with Neo4j plus an in-memory store so graphs could be built, updated, and queried on the fly.
• Wired up an interactive Cytoscape.js front end so you can explore and interrogate the graph visually, not just read a JSON blob.
• Shipped the whole NLP-to-visualization loop end-to-end as a team of 4, under HackMIT's weekend clock.

Impact

• A complete, working product in a weekend — text in, an explorable, queryable graph out.
• Makes hidden structure visible: relationships you'd never spot in prose become something you can click through and reason about.
• Sharpened the instinct that matters most in a product engineer — scoping a real full-stack build down to what ships when the clock is the hard constraint.