Our Investment in TrueFoundry: Transforming AI Deployments at Scale
The architecture of AI-enabled applications has quickly matured. What started as a thin wrapper around a foundation model has transformed into a more complex system involving multiple models, external tools, vector databases, application frameworks, and cloud environments. In a blog post, the Berkeley Artificial Intelligence Research Lab delves into the reasons and challenges of building such systems, which they refer to as “Compound AI” systems. This complexity is further exacerbated by the fact that unlike traditional machine learning (ML), which primarily involves ML scientists creating models, generative AI (GenAI) applications are increasingly being built by a new persona - developers and AI engineers. Additionally, while traditional ML applications had long development cycles, GenAI applications benefit from shorter cycles due to the ubiquitous availability of generalized pre-trained models which can be applied to a more diverse range of applications.
Enter TrueFoundry. TrueFoundry offers a unified platform as a service (PaaS) that enables enterprises to build, deploy, and manage GenAI applications on cloud or on-prem infrastructure with speed, scalability, and security. With TrueFoundry, AI teams can manage the rapidly expanding suite of models, iterate on building and testing GenAI applications, control the complexities of compute at deployment, and ensure efficient operations through single-pane governance. It’s no surprise that TrueFoundry has seen a 4x year-over-year growth in its customer base.
Some key use cases where TrueFoundry is already delivering value include:
- NVIDIA faced challenges in maximizing GPU cluster utilization and improving user experience due to the high demand for their GPUs. By leveraging TrueFoundry, NVIDIA built a multi-agent system to optimize GPU utilization, improving both performance and user experience.
- Siemens Healthineers needed a tool to achieve repeatable deployment of open-source models across business units. TrueFoundry’s platform provides easier management of key infrastructure components such as vector databases, pipelines, and seamless switching of models when needed.
TrueFoundry was co-founded by three incredibly smart founders - Abhishek, Anuraag, and Nikunj, who’ve known each other for over 15 years since their time at IIT Kharagpur. Abhishek and Nikunj then spent many years at Meta building deep learning models and managing at-scale infrastructure, and through that experience, realized what enterprises lacked to be able to scalably build AI systems, which resulted in the inception of TrueFoundry.
TrueFoundry’s team is focused on pushing the limits of AI infrastructure management – simplifying complexity and unlocking speed and scalability for everyone. Looking ahead, their team envisions a world where thousands of AI agents seamlessly interact with humans to accomplish tasks. In this future, where AI is embedded in every application, humans should no longer be the bottleneck. Instead, TrueFoundry sees a central platform where AI manages AI–a self-sustaining ecosystem that runs on autopilot, allowing organizations to focus on innovation, not infrastructure.
We are excited to be leading TrueFoundry’s Series A round with participation from our friends at PeakXV (formerly Sequoia Capital India), Eniac Ventures, and Jump Capital, and several angel investors including Gokul Rajaram, Mohit Aron, Cyan Banister, and executives from Fortune 1000s.
