Nomi_echo

💡 Inspiration

I’ve seen AI growing rapidly in the last few months. Everyone is launching tools to be more "productive", to write code faster or automate tasks. But I felt something was missing: Learning.

I didn't find any good use cases for "Human Acceleration." I believe AI shouldn't just do the work for us; it should help us learn faster. I don't want to just copy-paste code; I want to understand it. That’s why I started this project. I wanted to create a way to consume complex information (like documentation or manuals) quickly, not by reading, but by interacting.

🛠️ What it does

Nomi echo is an agent that helps you learn from any content. You simply upload your context a PDF, text, or documentation, and the agent helps you in three ways:

Audio Brief: It creates a short summary (like a podcast intro) so you can get the main ideas immediately.

Generate Questions: The agent quizzes you on the content.

Evaluate Answers: You can answer with your voice, and the agent listens and evaluates if you understood the subject correctly.

It’s perfect for onboarding new team members, explaining processes, or studying for certifications (like ISO 9001) in a short period of time.

⚙️ How we built it

I built this using Gemini as the core brain because of its ability to handle large contexts.

No Hallucinations: A key part of the build was reliability. By forcing the agent/MCP to search for answers only within the provided documents, I significantly reduced hallucinations. The agent doesn't guess or make things up; it sticks strictly to the source material you uploaded.

Audio Stack: I implemented Text-to-Speech and Speech-to-Text to make the experience feel like a real conversation in real-time.

MCPs (Model Context Protocol): I used MCPs to structure how the agent connects with the data.

Challenges we ran into

The biggest challenge was the time constraint. I had to learn about MCPs from scratch while I was building the app.

Accomplishments that we're proud of

I am most proud of building something that is actually useful to people right now.

Live Implementation: I currently have a mini-app running that is helping users learn about DeFi concepts in an interactive way.

Real Use Cases: The idea has sparked interest beyond just tech. I have friends wanting to use it to practice English (conversationally), and another friend sees value in using it to onboard new employees at their company.

Impact: It feels great to see that the tool is versatile and that I’m building something that solves real problems for real people, not just a tech demo.

What we learned

I learned a lot about integrating voice with AI. It’s not just about sending text; it’s about making the interaction feel natural. I also realized that converting static documents into a "conversation" is the fastest way to learn something new without worrying about false information.

What's next for Nomi_echo

I also wanted to explore a "pay-per-request" model (experimental), so I researched how to implement usage-based access so people only pay for what they use. And I want to improve the response time.

Built With

• fastapi
• gemini
• neondb
• typescript