SAN FRANCISCO — March 5, 2025 — Ceramic.ai emerged from stealth at the moment with software program for basis mannequin coaching infrastructure designed to allow enterprises to construct and fine-tune generative AI fashions extra effectively.
Based by Anna Patterson, former Google VP of Engineering and Gradient Ventures founder, Ceramic.ai stated it improves AI mannequin coaching pace and cost-efficiency, providing as much as a 2.5x efficiency enhance, accelerated by NVIDIA.
Ceramic.ai additionally stated it secured $12 million in seed funding from NEA, IBM, Samsung Subsequent, Earthshot Ventures and Alumni Ventures.
“Within the midst of a surge in AI adoption, too many firms are nonetheless hindered by obstacles to scale – from prohibitive prices to restricted infrastructure,” stated Patterson, CEO. “We’re democratizing entry to high-performance AI infrastructure so firms can navigate the complexity of AI coaching with out spending a whole lot of tens of millions in analysis and engineering assets. However the shift to enterprise AI isn’t nearly higher instruments – it’s about altering how companies work. If AI adoption had been a
baseball recreation, we’d nonetheless be singing the nationwide anthem.”
International AI investments are experiencing explosive development from $16 billion in 2023 to an estimated $143 billion by 2027. Regardless of this surge in spending, 74 p.c of firms nonetheless wrestle to scale AI successfully and obtain worth. A significant problem is that constructing AI infrastructure is dear, complicated, and resource-intensive. Whereas tech giants spend billions growing proprietary AI infrastructure, most enterprises lack the engineering assets to optimize and scale their very own AI fashions.
Present AI infrastructure can scale as much as 10x, however not 100x—true exponential development calls for an entire redesign. Ceramic.ai bridges this hole by offering an enterprise-ready platform that isn’t simply quicker however essentially extra scalable to energy the following technology of AI, dramatically decreasing the complexity and price of AI mannequin coaching.
The software program platform’s mannequin can practice with lengthy contexts and any cluster measurement, enabling enterprises to develop, practice, and scale their very own AI fashions quicker than conventional strategies. For smaller fashions, Ceramic.ai is as much as 2.5x quicker on NVIDIA H100 GPUs than present state-of-the-art platforms, and for large-scale long-context fashions, Ceramic.ai is the one viable selection for quick coaching.
Ceramic.ai has developed a complete platform that addresses the core challenges of enterprise AI deployment:
● Velocity and Effectivity: Ceramic.ai’s coaching infrastructure delivers as much as 2.5x larger effectivity than open-source stacks, chopping down coaching prices whereas enhancing mannequin efficiency.
● Unique Lengthy-Context Coaching Functionality: Ceramic.ai is the one platform that may practice giant fashions on long-context knowledge, offering unequalled high quality and efficiency. The corporate outperforms all reported benchmarks for long-context mannequin coaching, sustaining excessive effectivity even for 70B+ parameter fashions.
● Superior Reasoning Mannequin Efficiency: Ceramic skilled a reasoning mannequin for problem-solving and achieved an actual match Cross@1 rating on GSM8K of 92% tuning Meta’s Llama70B 3.3 base mannequin up from 78% and outperforming DeepSeek’s R1 84%.
● Optimized Knowledge Processing: Ceramic re-orders coaching knowledge, making certain every micro-batch is aligned by subject. Present approaches both masks away different paperwork, shedding the advantage of longer context size or take note of irrelevant paperwork, studying dangerous habits. By re-ordering coaching knowledge so it is available in 64k or 128k contexts, all on the identical subject, we improve the variety of knowledge factors the place consideration can be taught quadratically.
Constructed by a workforce of consultants in large-scale infrastructure, Ceramic.ai has already helped present enterprises cut back prices and enhance mannequin coaching effectivity in early trials. They’re partnering with Lambda, AWS and others for accelerated coaching.
“Ceramic.ai is a game-changer for AI builders and enterprises in search of elevated effectivity and superior price-performance,” stated Sam Khosroshahi, VP, BD & Strategic Pursuits I AI & Machine Studying at Lambda. “Mixed, our choices present prospects with an accelerated full-stack resolution, validated and backed by each infrastructure and mannequin experience. This permits prospects to attain quicker outcomes, decreased growth prices, and higher-quality options.”
“AI’s meteoric ascent has been like a rocket tethered to a horse-drawn carriage – till now,” stated Lila Tretikov, Companion and Head of AI Technique, NEA. “Anna and her workforce at Ceramic.ai have algorithmically shattered a crucial bottleneck in mannequin coaching, making it quicker, extra environment friendly, and scalable. With Ceramic, firms can scale their already large AI coaching workloads 100x – with out the corresponding surge in value or complexity.”
“Our funding in Ceramic demonstrates how IBM drives innovation and solidifies partnerships in extremely strategic areas,” stated Emily Fontaine, Vice President, IBM International Head of Enterprise Capital. “We’re thrilled to collaborate with Ceramic to deal with a crucial want to cut back AI compute prices, making coaching extra environment friendly and accessible.”
