Python Developer: Build Digital Audio Stream Harvester + Local Whisper AI (Two-Tier Output Gateway)

We are looking for a skilled Python Developer / Backend Automation Engineer to build a high-speed, completely digital audio harvesting and speech-to-text pipeline. Our target market is the New York City metropolitan area. To eliminate the need for complex physical radio antennas, line-of-sight range issues, and expensive 24/7 cloud AI bills, we are building a purely digital software-in-the-loop gateway. The software will connect directly to open-source public online scanner networks (such as OpenMHz or Broadcastify APIs), pull down raw New York City Fire Department transmission audio blocks into a local machine, process them 100% offline using a locally hosted Whisper AI engine, and isolate outputs based on a Two-Tier Architecture. *Note: You are primarily responsible for building the local ingestion, transcription engine, and outbound webhook/API delivery. We will handle all downstream CRM connections, n8n orchestrations, and advanced database workflows internally. Of course, if you are capable of handling the rest of the full-stack system integration, please feel free to quote us accordingly. * Technical Specifications & Scope: The script will programmatically track and harvest digital .mp3/.m4a/.ogg audio files as they are published online for the primary emergency radio feeds used across the New York City region. The developer will configure the system to monitor a total of 7 distinct channels: 5 Dedicated Borough Dispatch Channels (Manhattan, Brooklyn, the Bronx, Queens, and Staten Island) 2 Operational Citywide Channels Core Architecture (The Two-Tier Framework): Your software must handle incoming audio snippets using a parallel two-tier processing framework: Tier 1: The Master Slack Firehose (Human Intelligence Channel) Action: 100% of all harvested audio blocks from all channels must be transcribed locally. Delivery: Send a cleanly formatted message to a centralized human Slack logging channel (#fdny-master-logs). Payload: Must print the timestamp, the specific channel/borough name, the raw text transcript, and automatically attach/upload the compressed audio file to the Slack message for human reference. Tier 2: High-Speed Webhook Trigger (Automation Layer) Action: In parallel, a keyword scanner must scan every local transcript for severe operational fire escalation codes (e.g., "10-75", "All Hands", "Second Alarm", "Working Structure Fire"). Delivery: If a matching keyword sequence is found, immediately fire a structured JSON webhook payload directly to an internal endpoint we provide. Storage Management: The local machine needs a data retention policy. Raw audio files and local text logs must be stored locally for a short retention period (e.g., one week or one month, TBD) to act as a local backup archive. The script must include an automated cron job or cleanup routine to safely purge local assets once they cross this age threshold. Future Project Opportunities: This system is our foundational blueprint. Once stabilized in New York City, we plan on scaling this digital pipeline to ingest neighboring regional counties that utilize digital P25 trunked radio networks. The ideal freelancer will design this code modularly so adding new regions, channel counts, or digital streaming endpoints in the future takes minimal configuration changes. Required Skills: Strong proficiency in Python (specifically handling HTTP streams, async downloads, and JSON webhooks). Experience with Docker and Docker-Compose to containerize the local software stack. Hands-on experience deploying open-source AI speech-to-text models locally (such as faster-whisper or distil-whisper) running on consumer-grade CPU/RAM. Familiarity interacting with the Slack Web API (sending structured messages and programmatically uploading file assets). To Apply: Please briefly explain how you would structure a script to capture internet audio buffers cleanly without blocking, and which local Whisper engine implementation you prefer for maximizing transcription speeds on basic Linux hardware.