A startup situated in Palo Alto named Lamini has garnered $25 million in funding to enhance its platform designed for enterprises to deploy generative AI technology. Among the investors is a notable Stanford computer science professor, Andrew Ng.
The co-founders Sharon Zhou and Greg Diamos, established Lamini with a unique proposition in the market. They contend that numerous generative AI platforms are overly generalized and fail to provide infrastructure and solutions that align with corporate requirements. Conversely, Lamini has been meticulously crafted with businesses at the forefront, prioritizing high accuracy and scalability for generative AI applications.
Zhou, serving as the CEO of Lamini, emphasizes the significance of generative AI for organizational leadership, noting the challenges faced by companies in integrating this technology effectively. Despite the ease of creating working demos, transitioning to full-scale production often results in numerous setbacks.
Findings from a March poll conducted by MIT Insights reveal a gap between experimentation and adoption. While 75% of surveyed organizations have dabbled with generative AI, only 9% have embraced it on a broad scale. Obstacles include inadequate IT infrastructure, governance issues, skill shortages, and prohibitive costs. Security concerns are also prevalent, with a survey by Insight Enterprises showing that 38% of companies feel security issues hinder their ability to use generative AI technology.
Lamini’s response to these challenges involves an enterprise-centric optimization of their technology stack. This optimization encompasses both hardware and software, supporting model orchestration, fine-tuning, running, and training. A key innovation Zhou refers to is “memory tuning,” a training technique that enables a model to precisely recall specific data segments, potentially reducing erroneous outputs known as hallucinations.
While the concept of “memory tuning” is not widely recognized in academic literature, Lamini is confident in its benefits over other hallucination mitigation methods. Beyond this, Lamini’s platform boasts the ability to function in highly secure environments, such as those that are air-gapped. It offers the flexibility for companies to manage models across various configurations and can scale elastically to over 1,000 GPUs as required.
Lamini’s approach seeks to democratize control over generative AI, especially for enterprises concerned about the ownership of their proprietary data. The founders’ AI expertise and industry connections have played a crucial role in attracting investments. Notable backers include the CEOs of Figma, Dropbox, OpenAI, and LVMH, as well as AMD Ventures, First Round Capital, and Amplify Partners. AMD, in particular, has provided critical data center hardware support, allowing Lamini to run its models on AMD Instinct GPUs.
Having completed seed and Series A funding rounds, with Amplify leading the Series A, Lamini intends to use the $25 million to expand its team, enhance compute infrastructure, and delve into deeper technical advancements. As Lamini competes with enterprise AI offerings from tech giants, it has already secured paying users, including AMD, AngelList, NordicTrack, and various government entities. Despite the tech industry’s slowdown, Lamini experiences rapid growth, focusing on customer service with solid financials.
Amplify’s general partner recognizes the vast opportunity that generative AI presents for enterprises. Lamini stands out as a company genuinely addressing enterprise challenges with a solution that unlocks the value of private data while adhering to stringent security and compliance standards.
