Project Perseus | Speech & Voice AI Analyst

About the role

You will own the end to end integrity of speech and voice datasets from raw collection to production ready annotation. You will partner closely with data scientists and engineers to translate ambiguous project requirements into structured labeling tasks that directly shape model behavior. This role demands that you exercise sound judgment on every file you touch, balancing speed with accuracy under tight production timelines. You will be responsible for identifying edge cases, documenting inconsistencies, and escalating issues that could impact downstream model performance. Your work will ensure that training data remains representative, balanced, and aligned with real world speech conditions. You will maintain strict quality standards across repetitive workflows, catching subtle errors that others might overlook. This position places you at the core of the data pipeline, where your diligence directly determines the reliability of AI systems in production. You will be expected to communicate clearly with cross functional stakeholders and adhere to precise guidelines without constant supervision.

Key facts

What you'll do

Execute daily annotation tasks on speech, voice, and language data with a high standard of accuracy and consistency.

Follow detailed labeling guidelines, applying them consistently across diverse audio sources and linguistic contexts.

Conduct rigorous quality checks on labeled outputs, identifying and documenting anomalies before data reaches production.

Collaborate with analysts and engineers to clarify requirements, resolve ambiguous scenarios, and refine labeling rules.

Monitor workload queues, prioritize tasks based on project deadlines, and maintain steady throughput without sacrificing precision.

Track data quality trends, reporting recurring issues that could affect model training or evaluation pipelines.

Support the maintenance of annotation tooling, providing feedback that improves usability and reduces manual errors.

Ensure strict compliance with data handling policies, safeguarding sensitive information and maintaining documentation integrity.

Act as a first line reviewer, catching mislabeled segments and ensuring alignment with project specific objectives.

Communicate blockers and edge cases to senior team members, contributing to the evolution of labeling best practices.

Perform repetitive workflows with focus and discipline, sustaining high throughput over extended work sessions.

Participate in data audits, verifying that datasets remain balanced, representative, and fit for purpose in production systems.

Maintain organized records of decisions, changes, and exceptions to support traceability and future debugging efforts.

Demonstrate ownership of assigned data streams, taking responsibility for the quality and completeness of deliverables.

Requirements

Must be at least 18 years of age, with legal right to work in the United States without sponsorship.

Located in or able to physically commute to one of the following cities: NYC, Seattle, Bellevue, Redmond, San Francisco, Sunnyvale, Burlingame, Austin, Los Angeles, Washington DC, Chicago, Boston.

Authorized to work in the United States on the day of hire, with no requirement for future sponsorship.

Available to work full time, 40 hours per week, during standard business hours in the assigned location.

Able to perform the essential functions of the role using unaided human capabilities, with or without reasonable accommodations.

Possess the attention to detail and consistency required for high volume, repetitive data labeling work.

Demonstrated ability to follow complex instructions, apply rules uniformly, and maintain output quality over long work cycles.

Willingness to work onsite in a collaborative office environment, engaging directly with tools, teams, and workflows on location.

Nice to have

Experience with speech, audio, or voice data annotation in previous roles.

Exposure to phonetic transcription, phonology, or linguistic analysis.

Basic familiarity with data pipelines, version control, or quality assurance processes.

Comfort working with structured annotation interfaces and tooling.

Practical notes

This is a 100% onsite position with no remote work options.

Hours are fixed at 40 hours per week during standard business hours.

Candidates must be located in or able to commute to one of the specified cities in California and the surrounding regions.

No visa sponsorship is available for this role.

Employment is offered as W2 Full-Time with pay rates between $26 and $28 per hour, depending on location and experience.