Toloka
Broad/Unclear specialization
Quick facts
- Company
- Toloka
- Service type
- Data annotation / AI training data
- Specialties
- Image, Text
- Hiring status
- Both: hires workers and takes vendor projects
- Website
- toloka.ai
- Careers
- https://toloka.ai/careers
- Profile last verified
- 2026-01-29
Application process overview
Toloka is a global data labeling and AI training data platform originally founded within Yandex in 2014 and restructured as an independent company in 2024 after Yandex's reorganization. It connects enterprise AI clients with a crowd of contributors worldwide for tasks ranging from simple labeling to expert-level generative AI training.
Key findings
Application Process: Sign up via toloka.ai or Toloka's worker portal; contributors create a profile, verify identity, and can browse available projects. Expert tracks require CVs and skills screening.<br><br>Assessments: Projects typically start with training tasks and qualification exams; ongoing quality is measured via golden sets, overlap, and skill scores that gate access to higher-paying work.<br><br>Job Types / Expertise: Image/video/audio annotation, search relevance, content moderation, SFT/RLHF for LLMs, and specialized STEM, coding, and multilingual expert projects.<br><br>Compensation: Per-task micro-payments for crowd work (often cents per task); expert/AI tutor work pays hourly rates typically in the USD 10-40 range depending on domain and language.<br><br>Flexibility: Fully remote, asynchronous, and self-scheduled; workers pick tasks as they appear.<br><br>Challenges / Concerns: Crowd pay can be low, task availability fluctuates, strict quality thresholds, and payment methods / regional availability have changed since the Yandex split.<br><br>Legitimacy: Legitimate, well-known platform used by major AI labs and enterprises; covered by mainstream tech press and academic research on crowdsourcing.
Conclusion
Toloka is one of the more established players in the annotation and AI training data space, now operating independently after its 2024 spin-off from Yandex. It offers a two-tier model of open crowd tasks plus curated expert programs for LLM training, making it a realistic option for both casual taskers and subject-matter experts. Pay varies widely by track, and workers should expect qualification tests and strict quality controls. Overall it is a credible platform but not a reliable full-time income source for most crowd workers.