Senior Ai Integration Engineer Audio Llm Pipeline

## About the Role The Mission AASIOM is building the RAKSHA engine—a Phase 1 MVP for an AI-powered emergency dispatch system on the Western Express Highway. We are cutting accident response times from 23 minutes to under 60 seconds. The Role We are hiring a Senior AI Integration Engineer for a 6-week sprint to own the "brain" of the pipeline. You will receive raw WhatsApp voice notes, normalize them, and extract structured crisis data (severity, victim count, vehicle type) using Whisper and GPT-4o-mini. ## Responsibilities - Architectural Constraints: - Audio Normalization: Highway audio is chaotic. You must use FFmpeg to normalize levels and filter noise before it hits Whisper to ensure high-confidence transcriptions. - Strict Structured Outputs: The LLM must output validated JSON 100% of the time. You must utilize OpenAI's structured outputs with strict Pydantic schemas. - Fail-Closed Logic: If the AI detects ambiguity, hallucinates, or the transcript confidence drops below a set threshold, your code must immediately throw an exception and route to a human operator. Zero silent automated failures. ## Requirements - Python - OpenAI API - Audio Processing - Machine Learning - JSON ## What we offer - Location: Remote (India) - Type: Contract - Compensation: ₹80k – ₹1.2L ## About the Company AASIOM is building the RAKSHA engine—a Phase 1 MVP for an AI-powered emergency dispatch system on the Western Express Highway. We are cutting accident response times from 23 minutes to under 60 seconds. ## To Apply Start your application with the word "RAKSHA". Tell me exactly how you design your prompt and validation layer to prevent the LLM from hallucinating a JSON payload if the inbound audio is just 5 seconds of pure highway background wind.