Full Build

Every call, on the record

”
Hitch have listened to our needs and been committed to understanding the reality OSCA will be used in. The work has evolved continuously based on what the tests showed, and together we have adjusted both the approach and the solution along the way. It has been a close collaboration where our operational experience and their technical expertise have driven the development forward.
Ardenity

Why it mattered

When the alarm goes off, everyone talks on the same channel, short and fast. The OSC, the On Scene Coordinator, has to lead the operation and keep track of what has been said at the same time, and if you miss a few seconds there is nowhere to go back to. Ardenity wanted a shared record with the entire radio traffic written down while the operation is running.

What we saw

Ardenity arrived with a clear picture of what they wanted to build and how. What none of us knew was how radio traffic behaves when you try to transcribe it. Noise, clipped phrases, call signs and nautical terms that no general-purpose speech engine knows. That showed in the very first test, and it shaped the rest of the project.

What we built

OSCA is an Android app the OSC keeps on a tablet next to the radio. It records the traffic in short segments that Whisper transcribes on Ardenity's own machines. A language model then reads the whole conversation. It recognises speakers by call sign, fixes mishearings and compiles the situation in the Mission Hub. That is where the units involved, casualties and decisions sit in table form, so whoever has no time for the log reads the table instead.

What changed

We tested against recordings from exercises, found where it broke and rebuilt. What we have now is not what we sketched at the start, and it does its job. Everything that was said is still there, time-stamped line by line. Miss a call and you read up instead of asking for a repeat, and when a new coordinator takes over, the whole picture is in the log.

What it looks like

The radio talks. The log writes itself.

This is the chain behind the scenes. The audio segment comes in, Whisper transcribes it, the filter lets it through, the speaker is identified by call sign, the line reaches the devices and the Mission Hub situation picture is recomputed. Nobody touches a keyboard.

osca-whisper worker — VHF channel 16

/ utterances

Simulated radio traffic

What we achieved

Impact

All radio traffic written down, as it happens.

Units, casualties and decisions compiled from the log.

Hand over with the whole log, not from memory.

”
It is rare for a project to be this serious and this rewarding at the same time. A log that has to be right when it counts, in the middle of radio noise and call signs. We have learned along the way and changed direction more than once. Every time it got better.
Erik
Hitch

From insight to execution

  • Started from Ardenity's idea, not from a blank page

  • Tested against real recordings from exercises and rebuilt after every test

  • Changed approach when reality disagreed, from one line at a time to the whole conversation

  • Built layered filters against hallucinations in noisy recordings

What we used

Tech Stack

Laravel
React Native
Livewire
Expo
Laravel
React Native

Listen.

Understand.

Act.