The rapid evolution of AI and machine learning has transformed industries, and at the heart of this transformation lies the power of cloud partnerships. These alliances have not only accelerated innovation but also enabled businesses to harness AI/ML technologies at unprecedented scale and efficiency. From unlocking new insights in data to overcoming challenges like scalability, security, and accessibility, cloud providers have become indispensable collaborators in shaping the future of AI.
In this interview, Akshay Ram, an accomplished leader in cloud infrastructure and AI/ML technologies, delves into how cloud partnerships accelerate data-driven applications, address deployment challenges, and set the stage for the next wave of breakthroughs in AI/ML. He also provides a forward-looking perspective on how organizations can maximize their investments in AI/ML by prioritizing key business values and aligning their strategies with emerging trends. Prepare to explore a compelling vision for the future of cloud-based AI/ML innovation, driven by collaboration and a relentless pursuit of excellence.
Akshay Ram
How have cloud partnerships transformed the landscape or AI and machine learning innovation, particularly regarding speed and efficiency?
Companies are looking to apply AI to their core business to improve their customer experience as a first-order benefit and drive up productivity or improve cost efficiency as a result of better customer experience. Partnering with a cloud provider is key especially as you get access to accelerators, pricing options, and most importantly a community of other customers who have applied these solutions at scale. Learning what not to do is equally important as learning what to do and working with a cloud provider who has done this at scale for numerous customers is the most critical part of the partnership.
Could you share specific examples of how these partnerships accelerate data-driven applications and help businesses unlock new insights from their data?
There are numerous examples of how AI/ML is leveraging the cloud. Notably, generative AI companies are using cloud platforms to develop their foundational models, while enterprises are utilizing the cloud to consume and fine-tune these generative AI models. This is achieved either through services provided by cloud providers or by hosting open-weight models on self-managed cloud infrastructure.
What role do cloud partnerships play in addressing common challenges in AI/ML deployment, such as data security, scalability, and accessibility?
- Scale: We are now at trillions of parameters and customers need tens of thousands of accelerators to train these models. Partnership with a cloud provider will get you access to this scale.
- Access to Accelerators with a wide selection: Cloud providers provide multiple options of accelerators – the most obvious of GPUs but also within each accelerator, there are options to choose between training and inference. This allows customers to choose based on what best fits their needs.
- Security: This is a top priority of cloud providers and by partnering with them you benefit from shared responsibility where infrastructure is managed by the cloud provider and you also share some of the responsibility depending on which service you use. Additionally, there are a lot of build in value-added security tools for Gen AI such as prompt injections or jailbreaks
- Ecosystem: With AI/ML applications you need data. Cloud partnerships provide easy access to port data to object storage, and managed solutions for RAG patterns (including managed vector databases). All these enable you to build rich experiences
How do you contribute to shaping the roadmap for joint AI/ML solutions, and what factors guide your strategic decision-making?
The obvious answer is customer feedback, but what makes AI/ML unique is the dynamic and ever-changing market for generative AI solutions. Navigating this landscape is both an art and a science. When evaluating requirements, there are two key approaches to consider:
1) By looking beyond the alphabet soup of AI technologies and focusing on first principles, you’ll find that the same core cloud infrastructure needs consistently emerge;cost-efficiency, scalability, developer experience, and security.
2) Prioritizing and sequencing these requirements is the real challenge. This involves anticipating where customers will be in two years and using current customer feedback as valuable input. Leveraging data forecasts on usage can also provide further clarity and help guide decision-making.
What are some recent breakthroughs or emerging trends in cloud-based AI/ML applications that you attribute to strong industry partnerships?
The mass adoption of accelerators at scale would not have been possible without the cloud, enabling the training of models with up to 2 trillion parameters. Additionally, data repositories have scaled to petabytes, and the cloud provides the most efficient low-cost object storage solutions. By meeting customers where they are and leveraging their existing investments in the cloud, organizations can amplify the impact of AI/ML initiatives, aligning with and doubling down on a rapidly growing trend.
For organizations considering cloud partnerships, what key business values should they prioritize to maximize their investment in AI and ML?
Customer experience is enhanced by integrating AI/ML into products, with success often measured through developer experience. While cost and productivity are frequently discussed, they hold little value if the new AI-powered experiences fail to delight customers.
In terms of productivity, organizations are tracking metrics such as the contribution of new lines of code to AI development or the percentage of routine tasks automated by AI, which ultimately frees up employees to focus on higher-value work.
For cost efficiency, specific metrics vary across businesses, but they remain a critical factor in assessing the overall value of AI/ML initiatives.
Looking to the future, how do you envision cloud partnerships continuing to impact and evolve AI/ML technology, and what excites you most about this space?
I feel that partnering with the cloud will become the default implied action for AI/ML. However, there are some differences. Traditionally with web services and other non-AI/ML applications, there was always a brief talk about repatriation but over time, customers moved to the cloud and realized the benefits of lower capex, and more scale. However, a lot of these applications are something that customers migrated from on-prem (as a reference to compare), or even if they were new applications, there were already established best practices of running them at scale and efficiently. With AI/ML it is slightly different. A lot of these applications are newly deployed to the cloud (not migrated from on-prem). Gen AI applications themselves are new and a lot of best practices are still to be established. I’m most excited to see how this space will evolve as customers increasingly seek value-based outcomes. This shift will spark a vibrant phase of discovery, where best practices for optimization and scaling will emerge, driving innovation and unlocking new possibilities.
