SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Hippocratic AI is seeking an Audio Transcription Specialist to own the quality of transcription data powering their speech systems for healthcare AI. As the company scales voice agents across health systems, the quality of audio training data and the ability to explain ASR (automatic speech recognition) failures are critical to safe and rapid system improvement.
In this role, you will evaluate external transcription provider quality by reviewing transcriptions against annotation guidelines, delivering written feedback, and tracking recurring error patterns. You will improve quality processes by designing and refining QA workflows, review criteria, and transcription guidelines, then implement changes that close systematic gaps. You will validate audio data pipelines end-to-end to ensure training data is accurate and consistent with human judgments.
You will investigate production errors by analyzing audio and transcription failures in live ASR systems, documenting failure patterns, and assessing downstream impact. A key responsibility is building and maintaining an error taxonomy that researchers use to identify ASR weaknesses, prioritize improvements, and measure progress. You will collaborate across teams—presenting quality findings to researchers, engineers, and external providers, and partnering with them to translate findings into data and system improvements.
Within 90 days, you will stand up a repeatable QA process for external providers with refreshed review criteria and a first error taxonomy. Within 12 months, you will measurably reduce recurring transcription errors through provider feedback loops, validated pipelines, and a maintained taxonomy that guides ASR research.
The team moves fast with high autonomy, turning quality findings directly into model and data improvements. You will work closely with researchers and engineers building speech systems, and with external transcription providers.
REQUIREMENTS:
- Bachelor's degree in Linguistics, Speech and Language Sciences, or related field
- 2+ years of experience in transcription, linguistic annotation, audio quality assurance, or related field
- Strong listening skills and attention to detail; ability to identify subtle transcription errors
- Demonstrated ability to apply annotation guidelines consistently and exercise sound judgment in ambiguous cases
- Strong written communication skills with ability to explain errors and provide actionable feedback
- Comfort working with sensitive audio data and following data privacy and handling requirements
- Ability to work onsite in Menlo Park office five days per week
PREFERRED QUALIFICATIONS:
- Familiarity with ASR systems and evaluation metrics such as word error rate (WER)
- Experience evaluating speech across accents, dialects, noisy environments, or multiple languages
- Experience working with external annotation providers or research and engineering teams
- Experience identifying recurring quality issues and improving review processes
- Familiarity with spreadsheets, SQL, or scripting tools for analyzing quality trends
- Experience with healthcare audio or other privacy-sensitive, regulated data
