T

Toloka

AI · Amsterdam, Netherlands · Founded 2014

Toloka operates as a multinational data services company, headquartered in Amsterdam, providing human-in-the-loop data solutions essential for the development of artificial intelligence. Founded in 2014 by Olga Megorskaya, the company originated as a crowdsourcing platform for data markup to enhance machine learning algorithms. Megorskaya, who holds a degree in mathematical modeling, conceived of Toloka while working in the data team at Yandex, recognizing that in-house data labeling was inefficient. This insight led to the creation of a platform designed to solve the primary bottleneck in the AI industry: the scarcity of high-quality training data. Initially a part of the Russian technology enterprise Yandex, Toloka has since become a standalone entity. Following a significant corporate restructuring in 2023 and 2024, Yandex's Russian assets were sold, and its international divisions, including Toloka, were spun off into the newly formed, Netherlands-based Nebius Group. This separation has allowed Toloka to operate with greater independence and attract international investment. A major milestone occurred in May 2025, when the company secured a $72 million funding round led by Jeff Bezos's investment firm, Bezos Expeditions. As part of this investment, Mikhail Parakhin, the CTO of Shopify, joined Toloka's Board of Directors as Chairman. The firm's business model is centered around its platform, which connects AI developers with a global network of over 200,000 contributors, or "Tolokers," across more than 100 countries. These contributors perform a variety of microtasks—such as data annotation, content moderation, and sentiment analysis—that are critical for training, evaluating, and refining AI models. Toloka serves a client base of enterprise-level technology companies, including prominent names like Amazon, Microsoft, Anthropic, and Shopify, who use the platform to source high-quality, human-verified data for their AI projects. Revenue is generated by providing these data solutions to businesses. The platform has evolved from general crowdsourcing to offering specialized services, including PhD-level domain expertise for large language model (LLM) development and virtual environments for testing AI agents. Its services cover dataset creation, annotation, model evaluation, and safety testing through red-teaming. Keywords: data labeling, human-in-the-loop, AI data solutions, crowdsourcing, machine learning, data annotation, large language models, model evaluation, data-centric AI, generative AI, AI training data, red teaming, human intelligence, AI safety, natural language processing, computer vision, Olga Megorskaya, Bezos Expeditions, Nebius Group, AI agent testing, data quality, data infrastructure, human-augmented AI, dataset creation, AI alignment