Async

Inspiration

General-purpose AI tools are powerful, but they aren’t built around how students actually study. Reviewing long lectures, summarizing scattered notes, and organizing extended discussions often requires manual effort. We wanted to build something designed specifically for learning — an AI system that supports continuous thinking instead of one-shot prompts.

What it does

Async is an AI-powered review machine for students.

It supports real-time audio recording, automatic transcription, and intelligent summarization in a single workflow. Students can continuously send content, refine ideas, and generate structured notes without restarting the conversation or waiting for long processing cycles.

Structured note-generation modes, and developing long-context memory for large datasets.

Async turns scattered materials into organized knowledge.

How we built it

We built Async with a modular FastAPI backend, VUE frontend.

Real-time speech-to-text using Google Chirp 2

An OpenAI-compatible proxy for AI summarization

A pluggable transcription system (Google, Whisper, ElevenLabs)

Session management for persistent learning context

Server-Sent Events (SSE) for real-time streaming responses

The architecture is designed to be scalable and provider-agnostic.

Challenges we ran into

The biggest challenge was refining our system architecture while building new features. As requirements evolved, we repeatedly restructured modules to maintain clarity and scalability. Balancing long-term design quality with tight deadlines required constant trade-offs.

Accomplishments that we're proud of

Built a fully functional real-time AI learning system

Implemented stable streaming transcription and AI responses

Designed a flexible, pluggable provider architecture

Created persistent session-based learning workflows

We transformed a general AI interaction model into a structured learning-focused system.

What we learned

Building real-time AI systems requires strong architectural thinking, not just API calls. We gained experience in asynchronous design, streaming systems, and concurrency control.

We also learned that reducing latency and supporting iterative thinking significantly improves the learning experience.