AI Developer for Continuous Audio Transcription and Alert System

I am seeking a developer to build a simple personal-use AI automation that continuously monitors a fixed MP3 live emergency-services radio scanner stream and sends me automated incident summaries and alerts. The audio feed is from New South Wales, Australia (NSW). I will provide a fixed MP3 stream containing a mixture of NSW Ambulance and Fire and Rescue NSW (FRNSW) radio traffic. What the system needs to do * Continuously monitor the MP3 stream without any manual uploads. * Automatically transcribe the incoming radio audio using speech-to-text/AI. * Analyse the transcript and identify incidents from NSW Ambulance and FRNSW radio traffic. * Group related transmissions together so multiple radio calls about the same incident are treated as one evolving event rather than separate incidents. * Send me a concise AI-generated summary approximately every 10 minutes via Telegram or Discord. * Send an immediate alert when a potentially significant incident is detected, for example: * major fire * serious crash * CPR/cardiac arrest * stabbing * shooting * rescue/entrapment * multiple patients/casualties * other major emergency incidents * Run automatically 24/7 on a server/cloud service, without requiring my computer to remain on. Transcription accuracy Transcription accuracy is very important. The source audio is Australian emergency-services radio traffic and may include: * Australian accents * radio compression * background noise/static * clipped transmissions * overlapping voices * callsigns * abbreviations * emergency-service terminology * NSW street names and suburb names The developer should select and configure a speech-to-text solution that performs well with this type of audio. I can provide sample scanner recordings and example transcripts for testing and comparison. The system should also use recent context to help correct likely transcription errors, particularly NSW place names and common Ambulance/FRNSW terminology, rather than blindly treating every transcription as correct. Combined radio feed The current source is one continuous combined MP3 stream containing both NSW Ambulance and FRNSW radio traffic. The system should therefore attempt to determine which agency a transmission relates to based on terminology, radio codes, callsigns and surrounding context. Fire Rescue NSW radio calls mostly always begin with “firecom” and usually have the station number (“pump… 236”). I will provide reference material containing common NSW Ambulance and FRNSW radio codes, terminology and abbreviations. This material should be available to the AI as reference/context when interpreting and summarising transmissions. Where the meaning of a supplied radio code is known, the resulting alert should preferably describe the incident in useful plain English rather than simply repeating the radio code. Future expansion For the initial MVP, only one combined Ambulance/FRNSW MP3 stream needs to be monitored. However, the system should be structured so that additional MP3/live audio streams can be added later without rebuilding the entire system. This does not require a user interface or multi-stream dashboard at this stage. It is sufficient if additional streams can later be added through configuration/code and identified separately while feeding into the same incident detection and alerting process. Scope This is for my own personal use only. I do NOT need: * a mobile app * website * dashboard * public interface * user accounts * subscriptions I already have the MP3 stream and am looking for a simple, low-cost proof of concept/MVP, not a large production platform. Ideal experience includes: * Python * FFmpeg / audio streaming * speech-to-text * OpenAI or other LLM APIs * Telegram or Discord bots * backend automation * running lightweight services 24/7