Anna Patterson has enjoyed a remarkable journey in Silicon Valley. She has established three startups, notably including the search engines Xift and Cuil, as well as recall.archive.org, which evolved into the Internet Archive. Patterson served as the vice president of engineering at Google and later founded Gradient Ventures, a seed fund dedicated to artificial intelligence. Yet, her quest for innovation continues.
In a conversation with TechCrunch, Patterson expressed that she felt a strong desire to embark on a new venture a few years back, although she was uncertain if she had another startup in her. Following her breast cancer diagnosis in 2023 and subsequent time away from work, she came to a pivotal realization: she could either return to her previous life or initiate a new chapter.
Ultimately, she opted for the latter path: her latest startup, Ceramic.ai, aims to deliver infrastructure essential for foundational AI training, enabling companies to train large language models more efficiently using fewer GPUs compared to the “current state-of-the-art.” Ceramic claims its model can handle extensive contexts and operate with any clusters, aspiring to enhance model scaling by a factor of 100.
Patterson conceived the idea for Ceramic when she recognized that the existing framework for developing LLMs was overly complex and laden with variables, hindering widespread enterprise adoption.
“When you’re attempting to extend infrastructure tenfold, it typically goes smoothly,” Patterson explained. “However, when aiming for a hundredfold stretch or more, it’s often necessary to step back and reassess the situation. So, I reflected on whether this was the infrastructure I would want to deploy over the next decade.”
This line of thinking led to the founding of Ceramic AI in January 2024 in collaboration with Tom Costello, the chief scientist at the company. The startup has been operating discreetly since then, forming partnerships with organizations such as AWS and Lambda, although it has not yet commenced generating revenue. Patterson emphasized that she aimed to build credibility, awareness, and trust with potential clients before shifting her focus to sales.
Recently, Ceramic secured a $12 million seed funding round, spearheaded by NEA, with participation from IBM, Samsung Next, and Earthshot Ventures. Patterson indicated that NEA was an obvious choice to lead the funding round due to its technical expertise. The funds will primarily be allocated towards expanding sales and furthering product development.
After establishing three startups and holding senior roles throughout her career—alongside years spent as an investor—Patterson is aware that this journey feels different. She notes that her perspective on building startups has shifted since she transitioned to the other side of the table as a VC investor. Now, she senses a greater urgency as compared to when she initiated her earlier ventures.
“In my earlier years of starting companies, I experienced a rather relaxed phase where I was simply managing an engineering-focused, no-sales company,” she recounted. “Now, having spent years in venture capital, I definitely feel a heightened sense of time pressure.”
However, she considers this a positive development, believing that it’s advantageous to present your product to customers sooner, gather feedback, and iterate based on their input.
This insight is crucial for Ceramic, especially as it navigates an increasingly competitive landscape where other companies also seek to assist enterprises in scaling their foundational models.
One of Ceramic’s main competitors is Together AI, which aims to “turbocharge” model development and has secured over $530 million in venture funding. Another competitor, MosaicML, focused on accelerated LLM creation and raised $37 million prior to its acquisition by Databricks in 2023 for a staggering $1.3 billion.
At this point, Ceramic faces a significant challenge if it aims to carve out a niche in this rapidly evolving market.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
