Top companies hiring ASR & speech engineers in 2026

Hiring for speech recognition in 2026 is concentrated in four places: dedicated ASR-API companies, the big research labs, voice-agent startups that live or die on transcription quality, and vertical players in health, legal and automotive. Most roles are remote-friendly, and US total compensation typically runs from about $130K early-career to $300K+ at staff and research level (see our ASR engineer salary guide and Whisper salary guide).

This is a reference list, not a ranking — the "best" company depends on whether you want to train models, ship them to production, or embed them in a product. Live openings from most of these companies go into our weekly digest.

At a glance

Company	What they build	Typical speech roles	Base
Deepgram	ASR + voice-agent API, own models	Speech research, inference, ML eng	US / remote
AssemblyAI	Speech-to-text + audio-intelligence API	Applied ML, speech research	SF / remote
Speechmatics	Multilingual ASR engine	Speech modeling, ML infra	Cambridge UK / remote
OpenAI	Whisper, realtime speech API	Research, applied speech	SF
Google / DeepMind	USM, Chirp, on-device ASR	Research, software eng	Global
Amazon	Alexa, AWS Transcribe	Applied scientist, SDE	Multiple
Meta AI (FAIR)	wav2vec, MMS, Seamless	Research scientist / engineer	Menlo Park / remote
NVIDIA	NeMo, Riva, Parakeet/Canary models	Speech LLM eng, research	Global
Otter.ai	Meeting transcription	Applied scientist, speech	Mountain View
Suki / Abridge / DeepScribe	Ambient clinical documentation	ASR / ML eng	US / remote

Dedicated ASR / speech-to-text platforms

These companies sell transcription as the product, so nearly every ML, research and inference role touches speech directly.

• Deepgram — builds its own ASR and voice-agent models (the Nova family) and an API around them. Hires speech researchers, inference/optimization engineers, and ML engineers; strong on streaming and low latency.
• AssemblyAI — speech-to-text plus "audio intelligence" (summarization, topics, redaction). Applied-ML and research roles focused on model quality and new audio features.
• Speechmatics — a UK ASR company with a self-supervised engine covering a very wide language set. Speech-modeling and ML-infrastructure roles, often with UK visa sponsorship.
• Rev — large-scale transcription and captioning plus an API. Roles lean toward model operations, evaluation, and production ASR at volume.
• Gladia — a European ASR API company; smaller ML team, generalist speech/ML engineering.
• Otter.ai — meeting transcription and summarization at consumer scale; senior applied-scientist roles owning ASR and TTS end to end.

Big labs and platforms

The largest speech teams in the world. Expect a research track (publications at Interspeech / ICASSP / ASRU) and an engineering track shipping models into products used by hundreds of millions.

• OpenAI — Whisper and the realtime speech API; a small, senior speech team.
• Google / DeepMind — universal speech models (USM/Chirp), on-device recognition for Pixel and Android, and speech research across DeepMind.
• Amazon — Alexa and AWS Transcribe; one of the biggest ASR orgs, with applied-scientist and SDE roles across many sites.
• Apple — on-device speech recognition for Siri and dictation; privacy-constrained, efficiency-focused work.
• Microsoft — Azure AI Speech and the Nuance clinical/enterprise ASR business.
• Meta AI (FAIR) — wav2vec 2.0, MMS, and Seamless; foundational speech and speech-translation research.
• NVIDIA — the NeMo toolkit, Riva deployment stack, and the Parakeet/Canary model families; also speech-LLM and voice-agent roles.

Voice-agent and conversational-AI companies

ASR is upstream of everything these companies do, so speech-quality and latency work is first-class even when the job title says "ML engineer."

• Cartesia and Rime — TTS-led, but both hire for full-duplex and streaming speech pipelines.
• PolyAI, Vapi, Retell AI — voice agents for customer service; roles tuning and integrating ASR for telephony audio.
• Cresta, Observe.AI, Gong — contact-center and revenue intelligence; ASR plus diarization and analytics. (See also our speech analytics companies guide.)

Vertical ASR: health, legal, automotive

• Healthcare — Suki, Abridge, DeepScribe, and Microsoft/Nuance build ambient clinical documentation. The hard part is domain accuracy (drug names, abbreviations) and HIPAA-constrained pipelines. See medical ASR jobs.
• Legal — Verbit and Rev handle court and deposition transcription. See legal transcription jobs.
• Automotive — Cerence, SoundHound, and the automakers build in-car voice assistants with tight latency and noise constraints. See automotive voice AI jobs.

How to get hired

1. Pick a track. Research (novel models, publications), applied ML (fine-tuning, evaluation, data), or systems/inference (latency, cost, serving). The interview loops are different.
2. Ship something with real audio. A fine-tuned Whisper on a domain dataset, a streaming transcription demo, or a diarization + ASR pipeline with measured WER and DER beats a generic ML portfolio.
3. Know the metrics. WER, CER, RTF (real-time factor), DER, and how to read an evaluation that isn't leaking.
4. Contribute where the field lives. Whisper, faster-whisper, NeMo, icefall, SpeechBrain, pyannote — a merged PR gets noticed.
5. Show up. Interspeech and ICASSP are the recruiting venues; for 2026 that also means SLT 2026 in Palermo.