Ceramic.ai: Faster & Efficient AI Model Building | Anna Patterson

Anna Patterson's New Venture: Ceramic.ai

Anna Patterson boasts a remarkable history within Silicon Valley's tech landscape. She has been the driving force behind the creation of three distinct startups, notably Xift and Cuil – both innovative search engine initiatives – and recall.archive.org, which ultimately integrated with the Internet Archive. Her leadership extended to Google, where she served as Vice President of Engineering, and she subsequently launched Gradient Ventures, a seed fund specializing in Artificial Intelligence.

A Catalyst for Innovation

Patterson shared with TechCrunch that several years prior, she experienced a renewed desire to embark on a new entrepreneurial endeavor. However, she initially questioned her capacity for another startup. A breast cancer diagnosis in 2023, and the subsequent period of recovery, prompted a pivotal realization: she could either revert to her previous career path or forge a new one.

Introducing Ceramic.ai

The outcome of this contemplation is Ceramic.ai, a new startup focused on providing essential AI training infrastructure. The company aims to empower enterprises to accelerate the training of large language models, achieving faster results with a reduced reliance on GPUs compared to existing methods.

Core Technology and Scalability

Ceramic.ai asserts that its technology supports extended context windows and compatibility with diverse cluster configurations. A primary objective is to facilitate model scaling by a factor of 100.

Addressing Infrastructure Complexity

Patterson explained that the genesis of Ceramic.ai stemmed from her observation that current infrastructure for building LLMs is overly complex and presents too many variables for widespread enterprise implementation.

Rethinking Infrastructure for the Future

“While incremental improvements, such as a 10x performance boost, are often achievable,” Patterson stated, “significant scaling – 100x or more – necessitates a fundamental reassessment of the underlying infrastructure. I questioned whether the current approach would be optimal for the next decade of operation.”

Stealth Mode and Early Partnerships

Ceramic AI was officially founded in January 2024 by Patterson and Tom Costello, the company’s Chief Scientist. The startup operated in stealth mode initially, securing partnerships with key players like AWS and Lambda. Currently, the company is focused on building credibility and is not yet generating revenue.

Seed Funding and Strategic Investors

Recently, Ceramic.ai secured $12 million in seed funding, led by NEA, with participation from IBM, Samsung Next, and Earthshot Ventures. Patterson highlighted NEA’s strong technical expertise as a key factor in their selection as the lead investor. The funds will be allocated primarily to sales and ongoing product development.

A Shift in Startup Approach

Having transitioned from building startups to serving as a venture capital investor, Patterson notes a change in her approach. She now feels a greater sense of urgency compared to her earlier entrepreneurial experiences.

Prioritizing Customer Feedback

“In my earlier ventures, I enjoyed a more relaxed phase of focusing solely on engineering, without immediate sales pressures,” she explained. “However, my time as a VC has instilled a stronger sense of time sensitivity.”

Patterson believes this shift is beneficial, emphasizing the importance of early customer engagement and iterative development based on feedback.

Competitive Landscape

Ceramic.ai enters a competitive market, with other companies also focused on accelerating foundational model scaling.

Key Competitors

Notable competitors include Together AI, which has raised over $530 million in venture capital, and MosaicML, acquired by Databricks in 2023 for $1.3 billion after securing $37 million in funding.

Challenges and Opportunities

Ceramic.ai faces the challenge of establishing a distinct position within this rapidly evolving market. The company’s success will depend on its ability to deliver on its promises of speed, efficiency, and scalability for enterprise AI initiatives.