A short list this week, nine roles, with one theme running through most of them: speech is moving onto devices. Liquid AI is adapting a 1.5B-parameter speech-to-speech model to run on-device, Rivian and Volkswagen want an engineer to deploy speech recognition and wake-word models on a car's NPU, Starkey is researching DSP for AI hearing aids, and Meta is hiring research scientists for its glasses assistant. The top salary is a flat $300K from the Hankey Group for a senior voice-agent engineer in Los Angeles. For remote workers, Deepgram (US), Obble (worldwide) and Liquid AI all have options.
Research & modeling
- Member of Technical Staff, Post-Training, Applied (Audio) — Liquid AI
· San Francisco, Boston or remote (hybrid) · compensation not listed (base + equity)
Adapts LFM2.5-Audio, Liquid's 1.5B-parameter model that handles speech-to-speech, ASR and TTS in one network, for enterprise customers. The core problem is voice-driven function calling: turning a spoken request into the right API call with the right parameters, then confirming it back in speech. You'd own the data generation, SFT, preference alignment and RL, and the task-specific evals. Requires hands-on LLM post-training. Speech experience is a plus, not a requirement. View posting - Research Scientist, Wearables AI — Meta
· New York, NY + 5 other locations · $122K–$181K + bonus + equity
Speech and LLM research for the assistant on Ray-Ban Meta glasses and future wearables, in collaboration with Meta Superintelligence Labs. The work spans on-device and server models tuned for natural voice interaction. Requires a PhD, and you can apply while still finishing it. Preferred: large-scale ASR, TTS or LLM work and papers at ICASSP, Interspeech or NeurIPS. The pay range suggests an entry research level, so it's a realistic first industry job after a speech PhD. View posting - AI Communication Research Scientist — VinMotion (Vingroup)
· Cypress, CA · salary negotiable + performance bonus
Conversational AI for VinMotion's humanoid robots: dialogue generation, speech synthesis, Whisper-class ASR, and multimodal models that combine speech with vision (including lip reading) and gesture. Asks for a PhD or master's and 3+ years of research in generative AI, NLP or speech, with strong ASR and TTS skills. Publishing at ICASSP, ACL and NeurIPS is part of the job. VinMotion is the robotics arm of Vietnam's Vingroup. View posting
Voice agents & real-time systems
- Sr. AI Voice Engineer — Hankey Group
· Los Angeles, CA (onsite, no relocation) · $300K
Builds real-time voice-to-voice AI agents for customer service, sales and finance at the Los Angeles holding company behind Westlake Financial. The work covers STT/TTS integration, latency and streaming, telephony, and agent features such as memory, tool calling and RAG. 5+ years required. Note the stack: Python and C#/.NET, with Azure AI and Azure OpenAI listed as pluses. The posting lists a single figure rather than a range. View posting - Senior Full Stack Engineer, Solve Voice — Zendesk
· Austin, Madison or San Francisco (hybrid) · $145K–$217K + bonus
Builds the backend for Zendesk's real-time voice AI agent: telephony, WebRTC and audio integrations with outside speech and AI services, plus turn-taking, interruptions and handing calls off to humans smoothly. Some front-end work as needed. Requires 3+ years, Python or similar with async architecture, and real-time or event-driven systems (WebRTC, SIP, streaming APIs). SIP, media servers and ASR/TTS integration are pluses. The posting has been up 30+ days. Its page heading says "Senior Backend Software Engineer," but the description is for the full-stack role. View posting - AI Engineer, Voice and Realtime — Obble
· Remote (worldwide) · compensation not listed
Obble runs candidate screening interviews over live calls. This role owns the real-time interview worker (audio handling, STT, model calls and turn-taking within a latency budget), plus versioned prompts and an evaluation harness that has to prove a change made the interviewer better, not just different. Wants strong Python, async/streaming experience and LLMs shipped in production. WebRTC or LiveKit is a plus. It's a small company, and applying takes four fields and a resume. View posting
Training infrastructure
- Research Engineer, Machine Learning Systems — Deepgram
· Remote (US), Ann Arbor or San Francisco · $150K–$250K base + 10% bonus + equity
Works on the Research team to speed up how Deepgram trains its STT and TTS models: data preparation, high-throughput distributed training pipelines, automated evaluation, and orchestration and experiment tracking (Kubernetes, Prefect, MLflow). It also includes building internal tools that let non-technical colleagues use those systems. Requires hands-on experience with the ML research pipeline in STT or another speech domain, and data engineering for unstructured audio. View posting
Embedded, automotive & hearing
- Software Engineer, Edge AI — Rivian and Volkswagen Group Technologies
· Palo Alto, CA (hybrid) · $116.3K–$159.95K + 5% bonus + RSUs
An early-career C++/Rust role on in-vehicle software for Rivian R1/R2 and Volkswagen Group platforms. One of four possible starting projects is deploying speech recognition, TTS and wake-word models on the car's NPU against per-platform latency and accuracy targets. The others are telemetry, upload paths for ADAS and camera data, and the customer features built on them, so speech isn't guaranteed. Asks for 2–4 years or a strong project record. ONNX Runtime, TensorRT, TFLite or ExecuTorch, and VAD or wake-word experience are pluses. View posting - Sr. Signal Processing Research Engineer II — Starkey
· Eden Prairie, MN (onsite) · $124K–$189K + bonus
DSP and ML research for AI-enabled hearing aids in Starkey's Algorithms Group, taking algorithms from literature review and prototypes through lab and field evaluation into firmware requirements and products. Requires a master's in EE (or a PhD plus 3 years), with advanced DSP, acoustics and closed-loop acoustic control. MATLAB, Python and C/C++ are all expected. Psychoacoustics or cochlear-implant knowledge is a plus. The posting went up August 5, so it has been open a while. View posting
Hiring for a speech-tech role? Tell us about it and we'll include it here and in the digest. Browse everything by specialty on the homepage.