Waypoint EDU

Inspiration

College admissions has quietly become an industry that favors families with money. Private consultants can charge over $10,000, while most students are left navigating deadlines, requirements, and strategy alone. Even strong students miss opportunities simply because no one shows them how the system actually works.

We built Waypoint to level that playing field. Our goal wasn’t to make another chatbot, but to create a system that provides the kind of high-touch guidance students normally only get from private advisors. Strategy planning, essay feedback, and deadline management should not be a luxury.

What it Does

Waypoint is an AI-powered college application dashboard that actively guides students through the admissions process. It helps users plan application strategy, track deadlines, review essays, and get answers grounded in up-to-date admissions data. When the system encounters a school or requirement it doesn’t recognize, it can research reliable sources, update its knowledge, and improve for future users.

How We Built It

Waypoint is designed as a multi-agent system rather than a single AI model. • Backend (“Cortex”): Node.js and Express act as a central router that coordinates tasks and data flow. • Model Orchestration: • OpenAI GPT-4o handles conversation and student support. • Claude 3.5 Sonnet handles structured tasks like essay review, deadline auditing, and deeper research. • Voice Interface: Real-time voice conversations powered by ElevenLabs over WebSockets for low-latency interaction. • Infrastructure: Auth0 for authentication, with AWS and Supabase providing real-time relational data and document storage.

Each component is used intentionally based on the task, keeping the system fast, reliable, and scalable.

Challenges We Faced

Our biggest challenge came from combining Auth0 authentication with Supabase row-level security. Because users authenticated through Auth0 weren’t native database users, Supabase treated them as anonymous and blocked all read access. The frontend could write data but failed when trying to read it back.

To solve this, we introduced a secure backend proxy layer. All sensitive database queries now go through the server, which authenticates the user, fetches the data, and returns it safely. This resolved the access issues and improved overall security.

What We Learned • Action beats conversation: The real value of AI comes from doing work, not just answering questions. • Context matters: Supplying only relevant application state keeps responses accurate without unnecessary cost or latency. • Design builds trust: A polished interface dramatically improves user confidence and engagement.

Built With

• amazon-web-services
• auth0
• claude3.5
• css3
• express.js
• html5
• javascript
• node.js
• postgresql
• yutori