Applied Voice AI Engineer

About The Role

We are seeking a senior, full-stack ML generalist to lead an accent-tuned voice AI product from prototype to live production independently. You will architect and build the complete pipeline transforming raw voice notes into verified, structured instructions.

Key Responsibilities:

• Speech Recognition Model Development: Build an accent-tuned speech-to-text model with fine-tuning capabilities focused on high-risk keywords (digits, amounts, service names) and rigorous evaluation protocols.
• Data Extraction Layer: Develop systems converting transcripts into structured JSON instructions using LLMs with constrained output for precision.
• Voice Identity Matching: Implement speaker verification workflows including enrollment processes and one-to-one matching algorithms.
• Backend Orchestration & Security: Construct robust backends featuring idempotency, comprehensive audit trails, security hardening, and seamless API integrations with third-party payment gateways.
• Inference Infrastructure: Design self-hosted inference solutions serving models on in-region GPUs with high availability monitoring and uptime guarantees.

Tech Stack & Skills Required

• Demonstrable experience shipping ML systems end-to-end (beyond Jupyter notebooks).
• Expertise in Speech/ASR, specifically low-resource or accented audio scenarios.
• Solid LLM application skills: structured extraction, advanced prompt engineering, and model serving with vLLM or similar frameworks.
• Strong backend engineering proficiency: APIs, webhooks, third-party integrations, idempotency patterns, and logging best practices.
• Comfortable managing inference infrastructure including GPU serving, quantization strategies, deployment pipelines, and uptime management.
• A bias to ship fast from zero to a live prototype, followed by production hardening.
• Meticulous care regarding sensitive biometric data handling.