Tiny Fish raises $47M to turn the entire internet into a structured database for enterprise web agents
Aug 21, 2025 with Sudheesh Nair
Key Points
- Tiny Fish raises $47 million to build enterprise web agents that automate tasks across live websites by treating the open internet as a structured database.
- The startup's architecture snapshots live websites to generate training data rather than relying on synthetic replicas, with early customers including Google Hotels, DoorDash, and ClassPass running high-volume transaction automation.
- CEO Sudheesh Nair positions the $47 million as sufficient runway to control the company's trajectory as consumer browser agents, infrastructure providers, and code-generation tools converge around enterprise web access.
Summary
Tiny Fish raised $47 million in Series A funding to build enterprise web agents that automate browser-based work at scale. CEO Sudheesh Nair founded the company with co-founders from Meta and previously worked at Thoughtspot and Nutanix.
Enterprise knowledge work happens largely in browsers, but current AI models struggle with web interaction. Models that solve international math olympiad problems fail at basic web tasks because every click depends on context. Websites change, goals shift, pop-ups appear, and reasoning about seat selection at a movie theater requires understanding most people never explicitly encode. Tiny Fish believes it can automate this work and turn the open web into a structured, queryable database for enterprise customers.
How it works
Tiny Fish separates exploration from execution. Rather than training on synthetic websites or paying to replicate specific app workflows, the company snapshots live websites with proprietary architecture that generates synthetic training data from the actual web. Internal annotation and labeling pipelines process that data. At runtime, RL models handle understanding and intent, then hand off to code executed on CPU at scale. A variant of Google's AlphaEvolve improves the code over time.
The infrastructure layer handles volume browser spin-up, fingerprinting management, and anti-bot defeat.
Early customers
Google Hotels, DoorDash, and ClassPass are live customers. One unnamed customer manages 22 million web SKUs. Another operates 40,000 sub-shops. These deployments solve real enterprise problems while generating operational data that trains the system further.
Tiny Fish targets companies that run at enterprise scale, not individual transactions. Nair uses Expedia as the example: running 100,000 simulated transactions across 50 countries rather than booking a single hotel for one person.
Market structure
Three competitor groups exist today: consumer-facing browser agents like Perplexity's Comet and OpenAI's Operator, browser infrastructure providers, and code-generation tools developers use to connect software to the web. Nair expects all three to converge, with a handful of large companies controlling how internet access gets structured for enterprise use. He sized the $47 million raise to give Tiny Fish enough runway to control its own trajectory rather than face a forced exit or pivot.