Speech Algorithm Engineer (Speech LLM / SpeechLLM)$200,000 - $300,000San Francisco, Hybrid 3x per week in officeFull time / Permanent
About the Role
This company is already profitable, growing fast, and used by over 1.5M professionals globally. Revenue is tracking at ~$250M in under three years. The product works and is highly marketable, the next step is making its speech system significantly more accurate across languages, industries, and real-world conversations.
We're hiring a speech algorithm engineer to improve speaker diarization and keyword recognition in productio. This is applied, high-impact work that ships.
What You'll Do
Improve speaker diarization and multi-language speech recognition accuracy in real customer conversations
Design and optimize hotword and terminology recognition systems for industry-specific use cases
Fine-tune and train large speech models on substantial audio datasets
Build clear evaluation frameworks to measure keyword accuracy and speaker separation performance
Compare open-source and commercial ASR systems and push performance beyond them
Work closely with product and engineering to deploy models into live systems used dailyWhat "Great" Looks Like
You've trained or fine-tuned speech models on large-scale datasets (not small research-only projects)
You understand how speech systems behave in noisy, real-world conditions
You've improved measurable production metrics (accuracy, diarization quality, keyword recall)
You can read research and turn it into working systems
You take ownership when performance drops
Notable: If your experience is limited to light experimentation or purely academic research without production exposure, this likely won't be a fit.
Why Join
Profitable company at ~$250M run rate
Hybrid San Francisco team building both hardware and AI systems
Real ownership and visibility, not one engineer in a large org
Global product scale and meaningful datasets
Clear growth path toward senior technical leadership as the audio function expands
Strong data security and compliance standards, this is enterprise-grade infrastructure