AI Commerce Startup Thermit Raises Seed Round Led by Mashup Ventures
Mashup Ventures leads a seed round in Thermit, whose Creazy agent builds machine-readable product pages for 15,000+ sellers and hit monthly profitability in ten months.
By Nathan Brooks
3 min read

What's News
- Mashup Ventures led an undisclosed seed round in Korean AI commerce startup Thermit.
- Thermit signed more than 15,000 customers in 10 months after launch and reached monthly profitability.
- Creazy generates detail pages, thumbnails, banners and short-form videos in under three minutes.
- Creazy was trained on more than 50,000 pieces of sales-verified commerce content.
- Thermit says its AI multi-agent managing technology is the first in Korea with a patent application filed.
Korean AI commerce startup Thermit has closed a seed round led by Mashup Ventures, with 15,000-plus customers already on board just ten months after launch — and monthly profitability to match. The company declined to disclose the investment amount.
Thermit operates Creazy, a marketing content agent that analyzes a product's features and generates what online sellers need: AI-built product detail pages, thumbnails, banners and short-form videos. The whole process takes under three minutes. Thermit says Creazy runs on AI multi-agent managing technology, and that it's the first in Korea with a patent application filed for that approach — which the company says enables precise AI editing.
What makes the output different?
Here's the technical hook that separates Creazy from typical AI design tools. Detail pages made with Creazy are produced as code — HTML — that keeps the text intact. Existing AI design services usually output images, which machines can't parse. With Creazy, search engines and AI can read the text inside the content directly. According to Thermit, that helps with search visibility and with adapting to AI shopping agent environments.
The model behind Creazy was also trained on more than 50,000 pieces of commerce content whose sales performance has been verified. In other words, it wasn't learning from generic design examples — it was learning from content that actually converted. The system is designed to produce content that leads to purchases, not just content that looks finished.
How did it get traction without marketing spend?
Thermit signed its 15,000 customers through partnerships and word of mouth alone. No disclosed ad spend, no growth team mentioned — just distribution deals and referrals. The company also reached monthly profitability within that same ten-month window, a pace most SaaS startups don't hit until Series A or later.
Seungguk Lee, a partner at Mashup Ventures, said: "Thermit secured more than 15,000 customers through partnerships and word of mouth alone within 10 months of its official launch, and achieved monthly profitability." Lee said the firm valued the startup's quick access to sellers as the market moves toward an era of AI agents shopping on behalf of people. He also cited the founding team's expertise across development, product and sales.
Who's behind it?
CEO Jisung Kim previously led enterprise B2B AI software projects at a medical AI startup. He then founded a perfume brand and an AI multi-agent platform, building experience on both sides of the equation: running commerce operations and shipping AI products. The perfume brand gave him the founding insight. He found that producing marketing content took significant cost and time, and concluded that AI could bring a productivity shift to online commerce as well.
What does this change for sellers?
The practical question for merchants is what happens when AI agents, not people, do the shopping. Kim framed it directly: when AI agents find, compare and decide on products to buy, how accurately AI understands a product will determine sales. "That is why Creazy creates content as code that AI can read from the start," he said.
That's the strategic bet behind the seed round. If product discovery shifts from human browsers to machine agents, image-based listings become effectively invisible to the buyer. Sellers whose detail pages are machine-readable keep their shelf space; sellers whose pages aren't could lose it.
Kim added that Thermit plans to use the investment to build commerce infrastructure that helps sellers' products get chosen by both people and AI agents. For a ten-month-old company already profitable at 15,000 customers, the money goes toward infrastructure rather than survival — a signal of where Thermit thinks the commerce stack is heading.
Source: GN: Venture Capital
More from Nathan Brooks
Show full bio
News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.
1,577 articles