The Challenge
Emergency radio traffic carries the earliest signal of a developing wildfire — but only if someone is listening around the clock.
The Solution
Firefirefire transcribes emergency audio feeds with DeepSpeech and uses a Keras model to recognize the traffic that matters, triggering visual and audio alerts. Designed to run on a small embedded device with a software-defined-radio receiver.
What I Built
Automated Transcription
DeepSpeech turns live emergency audio into searchable text.
Alert Model
A Keras model flags the transmissions worth waking someone up for.
SDR Deployment
Runs on a small embedded device with a software-defined-radio receiver — no internet feed required.
Outcome
- Round-the-clock monitoring without a human in the loop
Built With