The data and talent layer behind frontier AI
HQ: Palo Alto, CA, United States | Founded: 2018 | Employees: ~500 (core team; plus a global talent network in the millions) | Stage: Series E — $111M at $2.2B (2025, led by Khazanah Nasional); profitable in 2024 | Website: https://www.turing.com
Turing supplies the human expertise that frontier AI runs on. Its network of millions of software engineers — originally assembled to fill remote-hiring gaps for tech companies — now does the specialised work of model improvement: writing hard coding problems and solutions, evaluating model outputs, red-teaming, and providing expert feedback in engineering and science domains. For AI labs racing to make models better at software, this is core R&D supply, not outsourcing.
The company was founded in 2018 by Jonathan Siddarth and Vijay Krishnan as a talent marketplace that used machine learning to vet and match remote developers. When generative AI took off, that vetting engine and engineer network became the raw material for something bigger — and Turing pivoted into what it now calls its AI data and model-improvement business, alongside its original developer staffing and AI-assisted engineering teams.
The financials reflect the pivot's success: Reuters reported revenue tripled to about $300 million in 2024, with the company profitable that year. In March 2025 Turing raised a $111M Series E led by Khazanah Nasional at a $2.2 billion valuation — a doubling in months — to expand its data operations as labs spend ever more on post-training.
Turing found the most valuable side-door in the AI boom. It started in 2018 as a straightforward idea — use software to match remote developers with companies — and built a network of millions of engineers. Then the AI labs came calling with a stranger request: they didn't just need coders to build products, they needed elite engineers to write the hard problems, examples and feedback that teach frontier models how to code.
That pivot made Turing one of the quiet suppliers behind models like those at OpenAI and other labs (TechCrunch has called it 'a key coding provider for OpenAI and other LLM producers'). Revenue tripled to $300 million in 2024 (Reuters) and the company reportedly reached profitability — rare in a category where rivals burn heavily.
Capital came on those terms: a $111M Series E led by Khazanah Nasional (Malaysia's sovereign fund) with WestBridge and others in March 2025, valuing Turing at $2.2 billion and taking total funding to roughly $250M. The company has since layered enterprise services on top: AI-assisted engineering teams for hire, and model-evaluation work as labs race on quality.
The risks sit with customer concentration and the strange politics of the data business. A handful of labs drive most revenue; when Meta bought 49% of Scale AI in 2025 and scrambled supplier relationships, the fragility of this supply chain became obvious. And as synthetic data improves, the value of human expert data must keep moving up the difficulty ladder — from labelling to the frontier-adjacent judgement only working engineers can supply.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki