Lean Progress IQ
DA NANG, VN  ·  ZRENJANIN, RS
AI evaluation work — for the rest of the world

Your passport is
the real application.

Every guide to AI training work is written for someone in Ohio. It never mentions that your ID country has to match your profile country, that half the platforms silently exclude you, or that your first payout can be frozen by a rule your bank never told you about. Start with what's actually open to you.

Eligibility below reflects published platform policy as of August 2026. Policies change — always confirm on the platform itself before uploading documents.

Select a country to see your real options.
Nine platforms checked against citizenship, ID requirements and payout rails.
Nine platforms · updated August 2026
Why this exists

I build the systems these platforms are grading.

I'm a Serbian citizen living in Da Nang. I hold a Serbian passport, a Vietnamese bank account, and a Payoneer balance — a combination that breaks about half the verification systems in this industry.

I also design AI agent systems. I wrote a 320-question evaluation benchmark, ran a 50-day longitudinal model trial publishing automated results daily, built fine-tuning pipelines on Vertex AI, and shipped a multi-agent framework to npm under MIT licence.

So when I say a rationale is weak or a rubric is ambiguous, that isn't a guess from someone who read a listicle. I've written the harness that grades them. And when I say a platform will reject your ID, it's because it rejected mine.

Most people entering this field are taught to label outputs. Almost nobody is taught what the evaluation is for. That gap is where the pay difference lives.

Where this comes from

  • aura-code — multi-agent framework, 1,559 tests, MIT licensed, published to npm
  • Aura Benchmark — 320-question agent evaluation harness; authored questions, scoring criteria and execution
  • 50-day trial — longitudinal model evaluation with daily automated published results
  • Fine-tuning pipelines — supervised fine-tuning on Google Vertex AI, ~11M-word Serbian legal corpus
  • Teaching — BA English Language & Literature, TEFL, 11+ years instructing corporate professionals
  • Enterprise operations — SAP ERP superuser, Lean/PDCA, automotive tier-1 logistics
What actually stops people

Four walls. None of them are about skill.

Almost everyone who fails at this fails before they ever see a task. Not because the work was too hard — because of an administrative wall nobody warned them about.

WALL 01

Citizenship vs. residence

If your profile says one country and your ID says another, verification fails — and most platforms give you two or three attempts total. Which country you declare is a decision to make deliberately, before you sign up, not a form field to fill in casually.

WALL 02

Silent geography gates

Several well-known platforms simply don't operate where you live, and won't say so until after you've built a profile. Others accept your signup but show no available tasks. Knowing which is which saves weeks.

WALL 03

Getting paid, actually

PayPal, Payoneer, Stripe and local banks each fail differently depending on where you are. Name spelling mismatches, foreign-currency restrictions, annual fees on low volume, and biometric thresholds can all hold your first payment.

WALL 04

Writing like an evaluator

Most applicants are screened out on a written sample. The bar isn't fluency — it's whether your rationale is specific, falsifiable, and useful to someone training a model. That's a learnable skill and almost nobody teaches it.

On price

Fair pricing, said out loud.

Plenty of companies already do this. They read your location, quietly adjust the number, and never mention it. I'd rather write it down and let you check the arithmetic.

The same hour of my time costs a different amount of life depending on where you're standing. Sixty dollars is an evening out in Berlin and most of a week's food in Lagos. Charging both people sixty dollars isn't neutral — it just hides the unfairness somewhere you can't see it.

So the price scales with your country's purchasing power, using published GDP-per-capita PPP figures. Same session, same preparation, same attention. Different number, for a stated reason.

One price for everyone is not the same thing as treating everyone equally.

I benefit from this too

I'm a Serbian citizen living in Da Nang. My own costs are low, which is precisely why I can price an hour in Nigeria at thirty dollars without it being charity. This isn't generosity — it's arithmetic that happens to work.

The floor is for me

I can earn around twenty-eight dollars an hour doing evaluation work directly. Below that, teaching costs me money rather than earning it. The floor isn't a sales tactic. It's the line where this stops being something I can keep doing.

No proof required

I won't ask for a utility bill, check your IP address, or make you prove anything. Tell me where you are and I'll take your word for it. If the adjusted number is still out of reach, write and say so — that conversation is not a problem.

What it costs

Priced against your economy, not mine.

Charging one flat rate worldwide would mean someone in Lagos pays four times as much of their income as someone in Berlin for the same hour. So prices scale with GDP per capita at purchasing power parity, floored so the work stays sustainable for me.

The formula is published below. You can check my arithmetic.

Select a country to see your rate.

How your price is calculated price = base_rate × √( your_country_PPP ÷ US_PPP ), floored Base rates are US-anchored. The square root compresses the spread — a straight linear ratio would put some countries under $10 and make the work impossible to sustain. The floor is my own opportunity cost: I can earn roughly $28/hr doing evaluation work directly, so no session is priced below that.

GDP per capita PPP figures: IMF World Economic Outlook and World Bank, 2024–2025 estimates in international dollars. Approximate and refreshed periodically. If your country isn't listed, ask — I'll calculate it the same way.
Start here

The eligibility & payments guide

A plain document covering all nine platforms: which accept your country, what ID they demand, how they pay, what each one actually pays, and the specific traps that block first payouts. Written from the outside-the-US position, because that's where I'm standing.

Free, no email required, no course upsell attached. Six pages. Take it and go.

Download the guide (PDF) Ask about 1:1 sessions
What I won't tell you I won't quote you a monthly income figure, because mine would be a sample size of one and yours depends on your country's rate tier, your available hours, and which platforms accept you. Anyone showing you screenshots of their earnings is selling something. This work is real, it is not passive, the task supply is inconsistent, and rates vary by where you live. I'll show you the published rates and the actual obstacles. What you make from that is yours to find out.