SuperAnnotate押注自有AI训练数据：3600万美元B轮融资，搜索量飙升1450%

做的是什么生意

San Francisco-based AI data platform: teams create, manage, annotate, fine-tune, and evaluate training datasets for computer vision and LLMs, with a marketplace of pre-qualified annotation teams.

启动资金：Early backing from KTH Holding (amount undisclosed); $36M Series B led by Socium Ventures (Nov 2024) brought total past $53M; ~$75M raised by 2026

起因

While pursuing a PhD on image segmentation at KTH, Vahan Petrosyan built a tool to manage and automate image annotation. He and his brother Tigran Petrosyan, an ETH Zurich graduate, along with KTH student Oscar Örnberg, went through KTH Innovation Launch from 2017, graduated in October 2019, and were admitted to UC Berkeley's SkyDeck accelerator. They founded SuperAnnotate to sell the tool to developers and enterprises.

经过

SuperAnnotate grew from image-labeling software into a data platform for fine-tuning, iterating, and evaluating AI datasets, adding RLHF workflows and a marketplace of pre-qualified annotation teams. Clients included Databricks and Canva. In November 2024 it raised a $36M Series B led by Socium Ventures with NVIDIA, Databricks Ventures, and Lionel Messi's Play Time Ventures, taking total funding past $53M; a later Series B-II brought it to about $75M. Kognic reported 5x software revenue growth in 2024 and a tripled customer base.

结果

Scaling: ~$75M raised with NVIDIA, Databricks, and Dell Technologies backing; 5x software revenue growth in 2024; tripled customer base; and a 1,450% search-growth month that made it #26 on Exploding Topics' August 2026 top-100 US trends list.

可借鉴之处

坚持变化中的切入点：图像标注工具演变为AI数据平台，关键在保持同一客户——构建训练数据的团队——当数据从图像转向大语言模型时。