Meridial is an AI training expert marketplace from Invisible. It recruits specialists for data, evaluation, and model-improvement projects. Rates and eligibility vary by project and location, so applicants should check the individual listing carefully.
Own call and ticket quality across two customer support teams serving a retirement savings platform. You will review agent interactions, audit existing QA reviews, and turn quality findings into coaching, training, and process updates. The role pays $8 per hour and runs 40 hours per week, from 9 a.m. to 6 p.m. ET. You will not take customer calls.
Record conversational customer support scenes in a feminine-sounding voice, performing both the customer and support representative. You’ll also review simulated cases in JSON, spot data or syntax problems, and explain fixes in plain English. To qualify, complete a JSON assessment and a brief language fluency assessment. Passing both does not guarantee a project placement.
As an independent contractor on an audio and language model evaluation project, you’ll assess Dutch samples for accuracy, fluency, naturalness, and overall quality. Your reviews may require close attention to grammar, pronunciation, tone, phrasing, and word choice, including whether spoken Dutch fits its context. Each task takes about 30 seconds to 4 minutes.
Evaluate AI-generated technical and customer-facing responses, drawing on language fluency, coding knowledge, and customer service judgment. You’ll check response quality and JSON structure against project guidelines. The available language cohorts are Dutch (Netherlands), Welsh, French (Canada), German (Germany), Italian (Italy), Portuguese (Brazil), Greek, English (US), and English (UK).
As an independent contractor, you’ll assess Danish audio and language model outputs for accuracy, fluency, naturalness, and quality. Your reviews will cover details such as grammar, pronunciation, tone, phrasing, word choice, and spoken Danish in context. Individual tasks take 30 seconds to 4 minutes.
Record spoken scripts in the project language, bringing clear pronunciation, conversational delivery, and expressive range to content such as phrases, numbers, dates, informational passages, and dialogue. You’ll follow recording guidance, respond to feedback, and deliver consistent, clean audio by project deadlines. The role is remote.
Evaluate AI voice and audio behavior inside a specified application version. You will follow detailed project instructions, record audio outputs, and assess how the system handles complex and changing test scenarios. The work focuses on instruction adherence, model logic, system behavior, and issues such as bugs, inconsistencies, or unnatural audio. German fluency as spoken in Germany is required, and the project will confirm device and operating system compatibility during application.
Use native-level Kansai fluency to support voice and audio evaluation projects. You will record spoken content, review audio samples, and compare recordings with written text, assessing accuracy, clarity, naturalness, and fit with project guidelines. The work calls for a strong ear for Kansai pronunciation, accents, phrasing, tone, and cultural nuance.
Review AI-generated coding and technical customer service content, including virtual agent replies, JSON structures, and audio-related material. You’ll check technical accuracy, clarity, usefulness, and Welsh language quality against project guidelines. The role requires native-level Welsh as used in Wales, but you do not need to live in the United Kingdom. Application assessments cover Welsh fluency and JSON skills.
Review product details and images for an online shopping and recommendation platform. For each listing, determine whether it belongs with an existing catalogue item, needs a new entry, or should be dismissed. The work also covers unusual cases, data inconsistencies, and suggestions for product categories and tags. A brief eligibility survey forms part of qualification, but completing it does not guarantee placement.
Help expand music source separation beyond the standard four-stem approach by identifying instruments and instrument families in short audio segments. The annotations support work on models such as HTDemucs, Open-Unmix, MDX-Net, and the RS-RoFormer family, with the goal of enabling more detailed remixing, music education, transcription, and music information retrieval.
As an independent contractor, you’ll support voice and audio evaluation projects in Cantonese as spoken in Hong Kong. You’ll record voice content, review audio, and compare spoken recordings with written text. The work calls for close attention to pronunciation, accents, phrasing, tone, cultural nuance, clarity, naturalness, accuracy, and project guidelines.
Location and eligibility
Remote · Eligibility not specified
Engagement type
Not specified
Compensation
$9/hr
cantonese
hong kong cantonese
audio evaluation
Language & LinguisticsView job
Pay
Pay on Meridial
Meridial pays project work biweekly. The platform says every project minute is paid.
Observed USD hourly rates: $6–$60 with a median of $11 across 21 current roles with a published hourly rate.
The summary uses the midpoint of a published range or the single published bound when only one is given.
Pay basis in current listings: 21 hourly, 5 fixed.
Checked Nov 19, 2026. Steps can change; confirm them on the platform before starting an assessment.
Questions
About Meridial
How many Meridial roles are listed now?
26 active roles are listed, and 26 of them publish compensation. Counts change as the source adds or closes roles.
Does Meridial pay hourly or per task?
The current listings use these pay bases: 21 hourly, 5 fixed. Meridial pays project work biweekly. The platform says every project minute is paid.
Which countries are represented?
The current listings do not specify country eligibility. Check the individual role before applying.
Are these Meridial roles remote?
26 of 26 active roles are remote according to the source. A remote role can still be limited to specific countries.
What happens after I apply?
Create a profile, apply to a project, complete the expertise review, and sign the project's statement of work. The full process, timeline, and official source link are in the application section above.