G.R.A.C.E

Inspiration

I’ve always found learning platforms too generic. I used YouTube, Udemy, Coursera, and others, but they all gave me the same content as everyone else. No matter my level or background, it felt like I was being handed a pile of information without a path. I wanted a tool that could break things down for me, adapt to how I think, and actually help me learn. This is what pushed me to work on G.R.A.C.E. Plus, I thought it'll be cool.

What it does

G.R.A.C.E. helps anyone learn anything by creating a step-by-step learning path. It breaks big topics into smaller parts and adds short tests after each one. Users can interact through voice, chat, or live video calls with an AI teacher. The platform adapts as you go and includes a public library where people can explore and share learning paths. It works for everything—from common subjects to niche, random skills.

How I built it

I used Gemini 2.5 Flash to generate the learning paths and checkpoint questions. ElevenLabs powers the voice interaction. Tavus handles live AI video tutoring. The frontend was built with React, and Supabase manages user data and content. I deployed the project on Netlify and plan to move to Cloudflare for better speed and scale.

Challenges I ran into

The biggest challenge was setting up the AI integrations. I thought things like Tavus and ElevenLabs would just work once plugged in, but I had to configure everything manually through their developer dashboards. Creating API keys, setting up projects, reading documentation—none of that was automatic. Once I got through that, things became much smoother. Other challenges included managing latency with voice and video, and keeping the interface simple even as the features grew.

Accomplishments that I'm proud of

I'm proud of building a real working product that users have already said helps them learn better. Some students told me they got value just from seeing how a topic was broken down, even without reading or clicking anything. That kind of feedback means a lot.

What I learned

I learned that structure is more important than content. People need guidance, not just information. I also learned that building AI tools takes more hands-on setup than I expected. And finally, I learned that small wins can give users a huge confidence boost.

What's next for G.R.A.C.E

We’re adding image generation to help explain ideas visually. We’ll keep improving adaptivity so the learning path changes based on how well someone is doing. And we’re expanding the public library so more users can find, remix, and share paths made by others.

We’re also adding tool use for AI tutors so they can go beyond talking and actually show things. This includes:

• Image generation: Tutors will be able to create visual explanations on the fly to help explain complex ideas better.
  
• Code blocks: For coding topics, AI tutors will generate and walk through code examples directly in the call or chat.
  
• Interactive responses: Tutors will soon highlight parts of a concept, generate diagrams, or pull in helpful references—all while teaching in real time.
  
• Smarter adaptivity: The system will update learning paths based on test results and user behavior, adjusting difficulty or focus without manual edits.

The goal is to turn G.R.A.C.E. into a hands-on, responsive learning system where the AI doesn’t just talk—it teaches with tools.

Built With

• javascript
• svelte
• typescript