AI Engineer, Agentic Voice (TTS)

Your Role
As an AI Engineer on our Speech Team, you'll own the back-end implementation and linguistic optimization of the voice (TTS) layer for our next-generation AI agents. You'll work squarely within our Speech Team, a high-impact R&D and engineering group focused on speech recognition, enhancement, and synthesis; bridging core speech science and product engineering so our agents sound human, context-aware, and trustworthy.

You’ll build the systems that render voice: integrating and optimizing TTS engines, engineering the persona and parameter machinery, and exposing voice attributes to our customer-facing UI. You'll partner closely with our Voice Experience Designer, who authors and owns the persona standards and quality bar that you implement — you own the platform that makes their designs real, fast, and consistent at scale.

This position reports to our Sr. Manager, AI Speech, is based at our Vancouver hub, and operates on a hybrid schedule.

What you’ll do

• TTS backend implementation: Own the integration and optimization of multiple TTS vendor APIs behind a unified interface with failover, and lead research and prototyping of open-source and in-house TTS architectures.
• Latency & pipeline engineering: Minimize time-to-first-audio and end-to-end latency across the ASR → LLM → TTS pipeline while maintaining or optimizing voice quality, in partnership with ASR and Audio AI engineers.
• Linguistic optimization: Apply your knowledge of phonetics and sociolinguistics to format TTS input for maximum naturalness — SSML tags, punctuation-driven prosody, and text normalization for names, numbers, dates, and currency.
• Persona system & parameter exposure: Build the persona parameterization system and architect the logic that exposes voice attributes to the product UI, implementing the house standards defined by Agent Experience Design.
• Prompt engineering as code: Manage structured LLM and TTS prompt templates with versioning and a rigorous evaluation harness.
• Conversational turn design: Engineer context- and state-aware "thinking" utterances that maintain caller trust while tool calls and model steps run under the hood.

Skills you’ll bring

• Technical foundation: Strong Python and hands-on experience with deep learning frameworks (e.g. PyTorch).
• Speech expertise: 3+ years in Speech Synthesis (TTS) or applied speech ML, including hands-on work with frameworks like NVIDIA NeMo, ESPnet, or Coqui, and with major TTS APIs such as ElevenLabs, Rime, and Cartesia.
• Linguistic background: Degree in Computational Linguistics, Computer Science, or AI/ML, with a strong understanding of phonetics, prosody, and syntax you can apply in code (G2P, pronunciation dictionaries, text normalization).
• Backend engineering: Experience building production-grade APIs and integrating multi-vendor services in a cloud environment (GCP preferred).
• Evaluation mindset: Fluency with speech-quality metrics (MOS, intelligibility, latency) and the ability to design and execute rigorous A/B tests for voice personas.
• Prompt engineering: Proven experience crafting, versioning, and evaluating LLM prompts (system, few-shot) and structured templates.

This is a builder role in a science/infrastructure org. If your strength is conversation design or prompt-tuning without production engineering, our Lead Voice Experience (VX) Designer
role is likely the better fit.