This position is no longer accepting applications.
Explore similar open roles below or browse all available jobs.
Multimodal Image Expert - Remote AI Image Rater
Overview
Moonlight seeks a Multimodal Image Expert to join the Image SxS Analysis project, where model names stay hidden and you judge pairs of AI-generated images created from one prompt. Each task asks you to compare two images against the source prompt, then decide which output succeeds on prompt alignment, visual quality, anatomy, structure, rendered text, and artifacts. You also record a short written rationale for every preference, including cases where both images look good or both miss the mark. The work runs on an hourly basis, starts with a paid onboarding quiz, and rewards steady, consistent decisions over rushed calls.
What You'll Do8
- 1Compare two AI-generated images side by side, using the original prompt as the benchmark.
- 2Score each pair for prompt alignment, visual quality, anatomy, structure, rendered text, and visible artifacts.
- 3Choose the stronger image, or label both as acceptable or both as flawed when neither wins.
- 4Write a brief rationale that explains the factors behind your preference.
- 5Keep model identities blinded throughout the review and avoid guessing which system produced an image.
- 6Apply the same standards across tasks so ratings stay consistent.
- 7Complete the paid onboarding quiz before moving to live project tasks.
- 8Follow project guidelines for edge cases, ambiguous prompts, and close calls.
Requirements6
- 1Fluent written English with clear grammar and readable explanations.
- 2Strong visual perception for spotting flaws in anatomy, structure, text, and image artifacts.
- 3Ability to judge images against a prompt and defend that judgment in a short written rationale.
- 4Consistent decision-making across repeated comparison tasks.
- 5Readiness to work on a remote, hourly project basis.
- 6No specialist degree required.
Who Should Apply
Someone with sharp visual instincts, fluent English writing, and patience for repeated image comparisons will fit this project well. The role suits candidates who can notice small differences in anatomy, rendered text, and artifacts, then explain those differences in a few clear sentences. People who rush through ratings or treat every image pair as a quick coin flip tend to score low, because the project values consistency and careful judgment over speed. Candidates also lose fit when their rationales stay vague, such as saying one image looks better without naming the prompt alignment or visual flaw behind the call. A specialist degree is not part of the bar, but weak written reasoning or poor attention to detail will block progress.
Salary Insight
The project pays $30/hr for image comparison tasks once you pass the paid onboarding quiz, which itself pays $20 regardless of the result. Compared with typical general data-labeling projects, this rate reflects the specialized nature of multimodal image review and the written rationale required for each task. The pay structure also separates the quiz from project work, so you receive $20 for the onboarding step even if you do not move forward. No salary range beyond the hourly rate appears in the source.
Location
Compensation
$30/hr
Required Skills
Application Tip
When you take the onboarding quiz, write rationales that name the exact prompt element, visual issue, or artifact behind each choice. A short, specific comparison gives reviewers evidence of your visual judgment, while a vague note like 'image A is better' can lower your fit score.
See NearSkill jobs more often in your search
Application & verification flow
1Instant rubric match
Your resume is scanned against this role’s requirements to check qualification fit.
2Screened before the employer sees it
Only profiles that clear screening are passed on.
3Outcome by email
We notify you at the address on your resume once the screening is reviewed.
Similar open positions
Explore active roles that match your skills and interests.
Turing
VerifiedKorean AI Analyst Photographer
Turing needs Korean-speaking photographers and annotators to build evaluation materials for multimodal AI agents. You will source authentic, licensed photographs of real-world places, products, and media from your locale, then mask, degrade, or crop them so key identifiers stay hidden. Each deliverable pairs a time-bound English prompt with grading criteria and solution notes, and agents must run multi-step research through Google Search, Google Lens, or reverse image search to recover the truth. The engagement runs as a contractor role for up to 16 weeks, with at least 4 hours of daily overlap with Pacific Time.
AfterQuery
VerifiedGraduate Student AI Evaluator
AfterQuery seeks current master's, PhD, MD, and MD/PhD students to join a remote AI evaluation research project across academic disciplines. You will use the vocabulary and visual conventions of your field, such as spectra, schematics, maps, or clinical images, to judge written and multimodal material. Projects run on a remote, asynchronous schedule, and you decide which windows to accept, with each window asking for about 10 to 20 hours per week. Pay ranges from $50 to $150 per hour, and a project week can bring in $800 to $1,500 based on discipline and seniority.
Micro1
VerifiedVisual Creative Expert
micro1 is building a team of Visual Creative Experts to work on a customer project that involves curating and evaluating aesthetic images. Your eye for composition, lighting, color, and originality will directly shape how AI models learn to judge visual quality. No background in AI is needed. What matters is your domain expertise in fields like photography, art direction, or graphic design. This is a remote contractor role where you review image collections and provide detailed written critiques.
Meridial
VerifiedImage Data Contributor and Evaluator
This project tests how well an AI-powered photo product reads and responds to pictures people contribute along with the context they attach. You send in eligible images and write comments or descriptions that explain what appears in each shot and what you know about it. Evaluators then compare the product's output against your notes to see if it handles image context, links details across photos, and produces results that feel accurate and safe. A short eligibility survey comes first, and a completed survey does not guarantee a spot.
Terac
VerifiedBrand Image Evaluator - Remote (US, CA, UK)
Terac runs a paid study on visual brand consistency and image selection. The work centers on how well each image matches established brand guidelines and which asset fits a brand's visual system. During a 30-minute calibration on a remote platform, you compare image pairs, pick the stronger match, and send feedback through a structured form. Finishing that exercise opens a standard onboarding path for future review tasks. A sharp eye for visual consistency matters more than a design portfolio here.
Turing
VerifiedImage/Video Annotator
Annotators support large multimodal AI systems by analyzing video and image annotation tasks, recording people, objects, environments, relationships, emotions, and storytelling flow from short clips and stills. You will describe how scenes change over time and note subtle visual cues like mood, lighting, and spatial relationships. The structured data you produce helps train AI models to perceive and reason about scenes, motion, and context similar to human understanding. Work is remote and based on a contract engagement with a 20-hour weekly commitment.