12 AI Training Platforms That Actually Hire and Pay PhDs in 2026

Something shifted in the AI labeling industry around 2024. Frontier labs ran out of easy gains from million-row Mechanical Turk annotations. The next jump in model quality had to come from a hundred radiologists rating cancer outputs, fifty patent attorneys building legal reasoning chains, or twenty mathematicians writing proofs the model could not paper over. A PhD became worth more than a clever screenshot prompt.

That shift is why this list exists. Pay at the bottom runs about $40 per hour for general expert work. At the top it reaches $250 per hour for narrow medical, legal, and frontier-research specialists. The 12 platforms below are the ones we have either applied to ourselves, watched contractors get paid by, or pulled first-party application data from over the last 18 months at RemoWork's company directory.

Who this guide is for

This list is for people who already have a credential the platforms below actively gate for. That means a PhD or ABD candidate, a postdoc, an MD, a JD, a senior research engineer with 5+ years in one specialty, or a published expert in a verifiable domain. If you are still in undergrad or learning your first programming language, the rates here are not going to land for you yet. Start with the platforms in our general AI training listicle, build a track record, and come back.

One honest note before we start. Many platforms below say "PhDs preferred" in marketing copy but accept master's holders or senior engineers with deep portfolios. If you are close to the credential bar, apply. The vetting process tells you the truth faster than the marketing page does.

How we ranked them

We score every platform on five things: verified top-end pay (with evidence, not marketing copy), realistic time-to-first-task after acceptance, how brutal the vetting is, payout reliability based on contributor reports, and the depth of the customer roster. The team behind this article maintains the company-directory side of RemoWork.life, which means we pull founder names, funding history, and hiring-page changes weekly. We have applied to seven of the 12 platforms below ourselves and rejected three from earlier drafts of this list because pay claims did not match what contributors actually saw.

The ranking is not strictly "highest pay first." A platform paying $250 per hour means little if you wait six months to get your first task. We weight the realistic onramp as heavily as the ceiling.

The 12 platforms ranked

1. AfterQuery: up to $250/hr for top-tier specialists

AfterQuery is the highest-leverage option on this list if you can clear vetting. Founded in January 2025 by three Wharton and Penn graduates (Spencer Mateega, ex-Silver Lake and Morgan Stanley and Meta and Google; Carlos Georgescu, ex-Citadel and Meta and Google; Danny Tang, ex-acquired-edtech-founder), AfterQuery closed a $30M Series A at a $300M valuation in April 2026, led by Altos Ventures with The Raine Group. Reports put annual revenue run-rate past $100M within 16 months of founding.

The network is roughly 100,000 verified professionals across finance, law, medicine, engineering, software, ML, business, and health. Pay tiers run $40 to $50 for general expert work, $80 to $150 for specialist tracks, and $150 to $250+ for the top tier. Apply at experts.afterquery.com. Typical onramp is 5 to 10 days for vetted candidates.

The catch: vetting is harder than the marketing page implies. Expect a domain-specific written task plus references.

2. Aligned Labs: OpenAI is a confirmed customer

Aligned Labs runs the contributor portal at alignedhq.ai with a network of about 5,000 experts from MIT, Stanford, Harvard, and other top institutions. Founded in 2024 in San Francisco by Paul McDonald (ex-Google), the company supplies frontier labs (OpenAI is a publicly confirmed customer) across 11 academic domains: biology, chemistry, computer science, economics, engineering, finance, law, mathematics, medicine, physics, and psychology.

Currently recruiting hard for physics, chemistry, biology, materials science, neuroscience, dermatology, radiology, ophthalmology, and otolaryngology. Aligned does not publish fixed hourly bands; contributors set their target rate during onboarding and the company matches you to projects in your range. Comparable specialist tracks at peer platforms pay $50 to $150+ per hour.

The catch: the rate-setting model means your first offer may be lower than you expect. Hold your number; if your domain is truly scarce they will come back.

3. Meridial AI: 24,000 specialists, backed by Invisible Tech

Meridial is the expert marketplace operated by Invisible Technologies. The network is roughly 24,000 vetted domain experts working on AI training for customers like Microsoft, Nasdaq, Cohere, AWS, and AI21 Labs. Compliance is unusually strong for a marketplace at this stage: SOC 2, HIPAA, and GDPR all in place, which makes Meridial one of the few options for medical and finance specialists who care about how their data flows.

Pay varies by track. Student and general AI trainer roles run $15 to $20 per hour. Entry-level freelance trainer roles cluster around $19 per hour. Specialist tracks (literature, anthropology, STEM, video production, legal reasoning, AI safety, creative writing) range $6 to $65 per hour. Domain mismatch is the most common reason for low offers, so list every specialty you can credential during onboarding.

Apply at meridial.ai/projects. Qualified experts can review project invitations within 48 hours of completing onboarding. Profile lives in the RemoWork directory.

4. Mercor: aggressive PhD recruitment at a $10B valuation

Mercor is the platform you have probably heard about because its three co-founders (Brendan Foody, Adarsh Hiremath, Surya Midha) became the world's youngest self-made billionaires at age 22. The company closed a Series D at a $10B valuation in 2025 and uses that capital aggressively to recruit PhDs. Pay for specialist tracks runs $20 to $100 per hour, with elite roles documented as high as $200 per hour.

The vetting flow is unusual. Mercor pulls signals from your LinkedIn, GitHub, and Google Scholar within minutes of you signing up, then schedules an AI-led video interview. If you pass, the time-to-first-task is among the fastest in the industry: often less than 72 hours. Read our Mercor review for the AI interview transcript.

The catch: the signal screen rejects more than 80% of applicants in the first pass, including some strong candidates. Your public credentials need to be detailed. A bare LinkedIn does not work here.

5. Outlier (by Scale AI): the largest PhD network on this list

Outlier is the contributor side of Scale AI. The network reportedly includes 700,000+ master's degree, PhD, and senior-credentialed contributors worldwide. Pay is tiered: general contributors $15 to $50 per hour, specialist tracks $30 to $100 per hour, PhD tracks $50 to $100+ per hour. Payouts go out weekly on Tuesday for work completed in the prior Sunday-to-Monday cycle, via Hyperwallet.

Outlier is the option for people who want predictable cadence over peak rate. The platform is mature, the payout pipeline is reliable, and task availability is more consistent than at faster-growing competitors. Read the Outlier review for the full task-availability breakdown.

The catch: task volume fluctuates aggressively by domain. A coding-track contributor may have 30 hours of work this week and three the next. Plan around it.

6. OpenTrain: published per-track rates and 15% flat platform fee

OpenTrain is the rare AI training marketplace that publishes its per-track rates publicly before you apply. Founder and CEO Weston Dotson built the company around contract transparency: a 15% flat platform fee, no hidden margin, and 152,000+ trainers across 180+ countries. Customer logos include MIT, Stanford, Harvard, Google DeepMind, Meta, and Goldman Sachs.

Published rates by track in mid-2026:

• Safety Red Team Analyst: $70 to $100 per hour
• Agentic AI Evaluation Lead: $65 to $95 per hour
• Code Generation Evaluator: $60 to $90 per hour
• Medical RLHF: $55 to $80 per hour
• Math Reasoning: $50 to $75 per hour
• Multilingual AI Trainer: $30 to $50 per hour

Apply at opentrain.ai/become-freelancer. The transparent-rate model also means you can negotiate a custom track if your specialty is not listed.

7. Pareto: Thiel Fellow founder, heavy Kenya footprint

Pareto is run by Phoebe Yao (Thiel Fellow, Forbes 30 Under 30). It pivoted to AI data labeling in 2023 and currently fields roughly 573 vetted trainers across six continents, with the largest single concentration (about 30%) in Kenya. The Africa-first hiring model is unusual on this list, and it means strong-but-undercredentialed candidates outside the US sometimes get accepted where American competitors would not.

Pay runs $5 to $50 per hour depending on track. The high end is reserved for verification work (the "verification layer for reinforcement learning on real-world expertise" is the company tagline) where domain credentials are non-negotiable. Funding is modest at $5.1M seed across MaC Venture Capital, Soma Capital, Jonathan Swanson, and others, so the company is more capital-disciplined than Mercor or AfterQuery.

The catch: average pay sits closer to the bottom of the range than the top for first-time contributors. Profile lives at /companies/pareto.

8. Mindrift: Toloka's premium specialist arm

Mindrift is the specialist marketplace operated by Toloka. The network is about 20,000 AI trainers, and crucially around 70% hold a master's degree or PhD. Coverage spans 90+ domains and 20+ languages. Pay sits at $15 to $100+ per hour. The top end is reserved for medical, legal, and senior STEM tracks.

Apply at mindrift.ai. The vetting includes a domain test that adapts to the specialty you select during signup. Mindrift is owned by Toloka AI AG in Switzerland, which is unusual jurisdictionally for an AI labeling company and matters if you care about EU data protection compliance.

Read our Mindrift review for the test breakdown. The catch: project availability is uneven. Some weeks are dense, others empty, and Toloka does not signal in advance.

9. SME Careers (SuperAnnotate): $20 to $130+/hr, weekly payouts

SME Careers is SuperAnnotate's expert recruiting brand. SuperAnnotate, founded in 2018 in Yerevan, Armenia, raised $50.5M total across Series A and B and is one of the few annotation platforms whose founders (the Petrosyan brothers plus two co-founders) all dropped out of PhD programs themselves. That cultural detail shows up in how the platform treats domain experts.

Pay is published openly at $20 to $130+ per hour. Payouts are weekly. The vetting is selective: comparable acceptance rate to Mercor's signal screen, but with a written task instead of an AI interview.

The catch: SuperAnnotate's main revenue comes from the SaaS annotation platform, so SME Careers is sometimes deprioritized when their enterprise pipeline is busy. Project availability mirrors their B2B sales calendar.

10. DataAnnotation.tech: $40+ per hour for STEM and PhD tracks

DataAnnotation is the contributor brand of Surge AI. Surge was founded in 2020 by Edwin Chen (ex-Google, ex-Meta) and has built a small empire of expert-labeling brands. DataAnnotation specifically is the writer and STEM-credential track.

Pay by track: general projects $25 to $30 per hour, multilingual $20 to $30 per hour, coding $50 to $75 per hour, STEM (master's or PhD credential gated) $40+ per hour. Glassdoor rates the contractor experience 4.0 from 158 reviewer reports. The reputation is solid, but the platform is notorious for unexplained account bans, so keep your work history outside the platform too.

Read our DataAnnotation review for the full story.

11. Defined.ai: 500+ languages, OpenAI customer

Defined.ai (formerly DefinedCrowd, rebranded 2021) is the option for linguists, speech specialists, and PhDs in language-adjacent fields. Founder and CEO Daniela Braga (PhD) built a marketplace of 1.6M+ contributors operating across 500+ languages and 150+ countries. Customers include Adobe, Amazon, Google, McAfee, Meta, and OpenAI. Funded at $63.4M total through Series B with Evolution Equity Partners leading.

Pay per Glassdoor analysis: $25 to $56 per hour depending on role. Defined.ai's contributor side feeds into the parent product (Neevo) and also into client projects directly. If your PhD is in linguistics, speech pathology, computational linguistics, or related domains, this is one of the few platforms where your specific training is the asset.

Profile lives at /companies/definedai. The catch: contributor payments to Neevo can take 1 to 2 weeks longer than competitors' weekly cycles. Plan cash flow accordingly.

12. Terac: $78/hr average across studies

Terac is the youngest platform on this list (founded 2025 in San Francisco by Zachary Baker and Jack Blair). Funded at $9M seed by Antigravity Capital. The network is roughly 100,000 vetted experts. Customers include market research clients like Maze, Strella, Tegus, Ballpark, getWhys, and Dialogue, plus AI labs running evaluation work.

Pay is unusual for the model: $78 per hour average across all studies. Typical participant tier runs $24 to $40 per hour. Senior specialist and deep-interview engagements pay $50 to $150 per hour. Trustpilot and Google both rate Terac 5.0 from real reviewer reports, which is rare in this category.

Apply at terac.com. The catch: study cadence is determined by client demand. A great profile may sit unused for a few weeks if your domain has no live studies. Profile at /companies/terac.

At-a-glance comparison

How to choose between them

Pick by your specialty first, by your priority second.

If you are a PhD in physics, chemistry, biology, neuroscience, materials science, or radiology, start with Aligned Labs and Mindrift. Both index hard for those domains and pay accordingly. If you are a JD, MD, or finance PhD, AfterQuery's top tier is realistically reachable. If you are a senior research engineer or ML PhD with strong public artifacts (papers, GitHub, model checkpoints), apply to Mercor first; their signal screen rewards exactly that.

If you care about cadence and predictability, Outlier and OpenTrain are the safer bets. If you want the highest possible per-hour rate and you are willing to wait through harder vetting, AfterQuery and SME Careers are the swing for the fences. If you are based outside the US or EU and worried about acceptance, Pareto's Africa-heavy footprint and Meridial's global compliance stack give you the best odds.

Apply to three or four in parallel. Vetting times are unrelated, projects do not overlap, and you give yourself optionality on which to prioritize once offers land.

How to actually get approved faster

Three things move the needle more than anything else.

First, your application essay should read like a job posting written in reverse. If the platform's customer is hiring "RLHF Specialists with experience in medical reasoning," you write your essay using that vocabulary. Match the entities (RLHF, chain-of-thought, evaluation rubric, edge cases) the way you would in a research proposal. The reviewers are pattern-matching at scale and your job is to be the easiest pattern to confirm.

Second, link your evidence. A bare statement that you have a PhD in biophysics is weaker than a Google Scholar link with three papers and a GitHub gist of a recent analysis. Even a personal Notion page works. The platforms want to compress vetting time and they pay people who make that easy.

Third, lead with the rarest credential first. If you have a JD plus a CS master's, lead with the JD; the JD pool is smaller. If you have a PhD plus fluency in Mandarin, lead with both, because the intersection is where the platforms pay specialists the top rate.

Frequently asked questions

Do I need a PhD, or is a master's enough?

A master's is enough for most platforms on this list, as long as your specialty is in demand. Mercor, OpenTrain, Mindrift, Outlier, SME Careers, DataAnnotation, and Defined.ai all accept master's holders. AfterQuery, Aligned Labs, and Meridial's top tiers tend to filter harder toward PhDs in narrow domains.

Are these 1099 contractor or W-2 roles?

All 12 are 1099 independent contractor arrangements for US-based contributors, with equivalent self-employed status outside the US. None offer health insurance, paid time off, or employer-side payroll tax contributions. Plan your tax withholdings accordingly. If you live in the US, expect to set aside roughly 25 to 30% for federal plus self-employment tax.

Do any of them require US citizenship?

None require US citizenship as a hard rule. Some specific projects (especially government-adjacent or US-defense-related work on AfterQuery, Aligned Labs, or Mercor) require US work authorization due to client compliance, but those are project-level constraints, not platform-level. International contributors get accepted on all 12; the only practical issue is payout method, where PayPal availability in your country can be the gating factor.

How much can a PhD realistically earn per month?

Based on contributor reports we have aggregated, a vetted specialist working 10 to 15 hours per week sustainably across two or three platforms earns roughly $2,000 to $5,000 per month. Top earners in narrow medical and legal specialties working full-time hours have reported $12,000 to $20,000 per month, but those are outliers, not the median. Treat anything above $8,000 per month as exceptional rather than expected.

Which platform pays the fastest?

Outlier (weekly on Tuesday via Hyperwallet) and Mercor (within 24 to 48 hours of task completion via Stripe Connect or similar) are the fastest. Defined.ai's Neevo side tends to be the slowest, with 10 to 14 day payment cycles. Most others pay biweekly or weekly.

The story to remember from this list is the 2024 shift. Frontier AI labs decided they could not climb the next quality curve with cheap labelers. They opened budgets for specialists. The 12 platforms above are the ones that built around that thesis early. If you have the credentials, you are in a rare moment where the demand side genuinely needs you, and the rates reflect it. Bookmark this page. We refresh it every 90 days as platforms launch new tracks and change rates.

Related reading: Top AI Training Platforms That Pay You to Train AI Models in 2026 · Pila8 Review · Mercor Review · Outlier AI Review