Skip to main content

FiftyFive Tech

Email Us sales@fiftyfivetech.io
Get in Touch
Smiling girl in glasses and a yellow shirt waving at a laptop during a video call, with a stack of books and a sheet of paper on the white desk in front of her

Generative AI applications

Real-time AI tutoring backend with intent routing and session memory

FiftyFive built a FastAPI and WebSocket AI tutoring backend with intent routing and session memory; educators reported 40% less repetitive work.

Discuss your real-time AI product
Industry
Education — online tutoring
Region
Solution
Real-time AI tutoring with intent routing
Story

About the client

The client delivers AI-based tutoring to students and institutions. Conventional e-learning platforms often offer little personalisation, respond slowly and give little insight into how students engage. The client wanted tutoring that felt closer to a live classroom, while taking repetitive work off educators.

Challenge

A tutor that answers instantly and remembers the conversation

  • Low latency. Students expect answers in real time, not after a page reload.
  • Many kinds of request. A question about a topic, a request for the next lesson and small talk each need a different response path.
  • Continuity. The tutor had to remember earlier exchanges to keep sessions personal.
  • Adaptive paths. Learning paths had to adapt to different students.
  • Scale and resilience. The system had to scale and tolerate faults as more institutions joined.
Solution

What FiftyFive built

A dedicated FiftyFive team covering backend, AI/ML and cloud designed, built and deployed the platform. The scope covered: a FastAPI backend, with asyncio for concurrent, asynchronous execution; WebSocket connections for low-latency, two-way communication; an intent recogniser that routes each user input to the right module; integration of AI/ML modules for question answering, course delivery and conversational support; a context engine that persists conversation history for personalisation; deployment on Google Cloud Run with automated scaling; local file storage and custom logging to track student interactions.

How the solution works

  1. Real-time connection

    • Each student session runs over a WebSocket connection to the FastAPI backend. asyncio lets the backend handle many sessions concurrently without blocking, which keeps responses fast as usage grows.
  2. Intent routing

    • Every message first reaches the intent recogniser, which decides what the student is asking for and routes it to the matching module: question answering, course delivery or conversational support.
  3. Session memory

    • The context engine stores the conversation history for each session.
    • Modules use it to answer in context, so the tutor builds on what the student has already asked and learned.
  4. Deployment and tracking

    • The backend runs on Google Cloud Run, which scales instances automatically with demand.
    • Local file storage and custom logging record student interactions so the team can track and improve them.

Technology stack

FastAPIBackend API framework
asyncioConcurrent, non-blocking request handling
WebSocketsLow-latency, two-way messaging with students
Intent recogniserRouting each input to the right AI module
Context enginePersisting conversation history per session
Google Cloud RunContainerised deployment with automated scaling

What the solution enabled

Outcomes

Business transformation

40%

Less repetitive work for educators

Real time

Tutoring served over WebSockets from Google Cloud Run

3

AI modules for question answering, course delivery or conversational support

40% less repetitive work for educators

As reported by educators using the platform.

Real-time tutoring in production

Served over WebSockets from Google Cloud Run.

Personalised sessions

With conversation history carried through each session.

Scales automatically

On Google Cloud Run with demand.

FiftyFive Technologies designed, built and deployed the backend of an AI tutoring platform for an online learning provider that wanted classroom-like interaction at scale. Built with FastAPI, asyncio and WebSockets on Google Cloud Run, the backend routes each student message through an intent recogniser to AI modules for question answering, course delivery or conversational support, and a context engine keeps conversation history so sessions stay personalised. Educators reported a 40% reduction in repetitive workload.

FAQ

Technical questions

How do you keep an AI tutor responsive in real time?

For an online learning provider, FiftyFive Technologies built the tutoring backend on FastAPI with asyncio and WebSockets, so each student keeps an open two-way connection and many sessions run concurrently. Google Cloud Run scales the backend automatically with demand.

How does the tutor know which kind of help a student needs?

An intent recogniser reads each message and routes it to a question-answering, course-delivery or conversational-support module. A context engine keeps the conversation history so each module answers in the context of the session.

Discuss your real-time AI learning product with FiftyFive's AI engineers.

Book a call

Project snapshot

Industry
Education — online tutoring
Services
Backend development, AI module integration, cloud deployment
Team
Backend, AI/ML and cloud engineers
Status
Deployed on Google Cloud Run
Core technologies
FastAPI, asyncio, WebSockets, Google Cloud Run