Artificial intelligence startup Hark has entered a multi-year strategic partnership with NVIDIA to develop personalised agentic AI, combining advanced computing infrastructure with deep technical collaboration. According to the company announcement, this partnership will provide Hark with gigawatt-scale compute capacity built on NVIDIA's next-generation Vera Rubin platform. The infrastructure will support AI training and inference as Hark prepares to launch its platform publicly before the end of summer. As reported by Hark, the partnership gives the company the foundation to build AI that "actually knows you, speaks your language, is highly personalized, and lives on hardware made for you."
Hark is building multimodal AI systems designed to use speech, vision and persistent memory to create a more personalised experience. The company aims to develop AI that can work across existing devices as well as its own bespoke hardware. As reported by Hark, the startup will use several NVIDIA technologies across its AI stack, including the Megatron stack for model training, Dynamo for inference, permissively licensed Nemotron pretraining data and NVSentinel for GPU cluster resiliency. NVIDIA's AI infrastructure has already supported the development of Hark Handoff, the company's computer-use model. The technology will also support distributed inference when Hark's platform launches and the future training of its own foundation models.
According to Hark, the company is developing its AI models, native hardware and new interfaces together, with its agentic systems designed to interact naturally with people and the real world. Hark founder and CEO Brett Adcock stated that the NVIDIA partnership would provide the infrastructure needed to develop highly personalised AI and deploy it quickly and at scale. The company will bring its initial AI platform to market soon, with NVIDIA's computing infrastructure supporting its training, inference and longer-term development plans. As reported by Hark, the company is developing the everyday interface of AI, building multimodal systems that remember who you are and what you plan to do. The company is developing products that work for everyone, with its agentic systems designed to interact naturally with people and the real world.
Nico Caprez, NVIDIA's vice president of Global AI Infrastructure Growth, emphasized the computing requirements for personalised AI development. As reported by Hark, Caprez stated that "Personalized AI that combines speech, vision and persistent memory requires enormous computing infrastructure across training and inference." The partnership will support distributed inference when Hark's platform launches and the future training of its own foundation models. By building on NVIDIA's full-stack accelerated computing platform at gigawatt scale, Hark will have the foundation to develop, train and run AI at scale. Building with NVIDIA provides the foundation to scale this generational technology.
