13 May 2025

CalliQ: runner-up at the SSE Business Lab Hackathon

Two CalliQ phone screens, an incoming call from Lotta and an in-call warning not to share bank account details over the phone.
Didrik Munther Didrik Munther

CalliQ: runner-up at the SSE Business Lab Hackathon

Two weeks of building and countless insights as Simon Wanna and I joined the SSE Business Lab × Microsoft × KTH AI Society Hackathon. We built CalliQ: an AI tool that stops fraud attempts in real time.

With more than 250 applications and 14 finalists, we were happy to finish as runner-up and earn a spot in SSE Business Lab's Activate program to keep on building. Activate was designed to help early-stage teams validate and build their business ideas.

How CalliQ works

CalliQ listens to phone calls in real time. When a caller seems to be attempting a scam, CalliQ warns the person receiving the call while they're still on the line.

We built three views: the caller in a browser, the receiver on a phone, and a dashboard that shows the transcript as it comes in. On stage at Microsoft in Stockholm, we demoed it with a live call between two devices. During a call:

  1. The caller's browser records the microphone, downsamples the audio to 16 kHz, and streams it over a WebSocket to our Python backend.
  2. The backend connects the caller to an available receiver and transcribes the audio with Azure AI Speech, building up a transcript as the conversation goes on.
  3. After each new sentence, an agent built with LangGraph and GPT-4o mini reads the conversation so far and decides whether the receiver is about to be scammed.
  4. If so, the agent calls a warning tool with a short explanation, and the backend sends that warning straight to the receiver's phone.

We asked the model to keep its warnings friendly and easy to understand. Instead of a generic alarm, the warning explains what looks risky:

It's not safe to share your bank account details with anyone over the phone, especially if they are asking for it unexpectedly. This could be a scam.

CalliQ's receiver view on a phone, showing an incoming call from Lotta with Decline and Accept buttons. CalliQ's in-call view 16 seconds into the call, warning: It's not safe to share your bank account details with anyone over the phone, especially if they are asking for it unexpectedly. This could be a scam.
The receiver's phone in our prototype: an incoming call, then the warning 16 seconds in. Open a screenshot for a closer look.

Big thanks to everyone from SSE Business Lab, Microsoft, and KTH AI Society who made this happen. If you're organizing a hackathon, Hugo Mårdh is the one who knows how to make it a success. You can read more about the event and the other finalists in SSE Business Lab's recap.

Looking forward to the next one. Cheers!