How LinkedIn’s platform strategy helps us innovate with speed and scale in this GenAI moment

Generative AI (GenAI) enables us to do tasks with greater ease than ever before – from writing first drafts of InMails to delivering intelligence to salespeople before they head into customer meetings. But delivering these easy, seamless experiences to our customers and members is incredibly complex, and it’s prompting us to think differently about the way we innovate.

The AI moment we’re in now reminds me of the mid 2010s, when the mobile boom was fundamentally changing the technology landscape. During this transformative moment, we had to develop a new technology stack and rewrite every line of mobile code to enable cutting edge mobile experiences for our customers. It was a challenging process, but it ultimately set the stage for the next ten years of product development and innovation at LinkedIn. As we're working to meet today's GenAI moment, we have a similar challenge in front of us. 

We’ve spent much of the past year reimagining our engineering playbook with an AI-first mindset, tackling the typical challenges that slow innovation in any organization. We want to make it easier for our developers to quickly build and ship new products and features. As part of this effort, we’re stepping up our platform strategy. 

Our platform strategy is key to building products at speed and scale. These investments remove roadblocks for engineers and developers so they can focus on what they enjoy most about their jobs - building incredible products that help professionals around the world. 

One way to think about this strategy is to imagine a company’s technology stack as a bustling city's urban landscape. There are roads and utilities as the foundations, marketplaces and public squares that facilitate commerce, business and the exchange of ideas, and at the highest level, stores and neighborhoods that drive connections in everyday lives. 

With our platform strategy, we're tackling that middle layer - making it easier for the city to build businesses on top of the foundations. We want our developers to quickly access the technology capabilities they need to build and foster great user experiences, faster. 

Streamlining the developer experience for our engineers 

Among the biggest roadblocks for our application developers is the time it takes them to discover what platform capabilities exist – and then figure out how to integrate with them. So, we put our focus on making this as simple and as close to self-serve as possible.

A great example of how we’ve reduced complexity for our developers over the past year is how we’ve made it easy for our teams to experiment, test, and launch GenAI products and features quickly. Here, we built out support for OpenAI models via Azure OpenAI Services, as well as internal and open-sourced models, and made it easy for developers to choose models based on the particular use case and goal for the GenAI product they are building.

We’ve also taken steps to adapt our internal systems for a GenAI world by building developer libraries to turn structured API responses into prompts. By extracting structured data into prompts, our GenAI-powered tools can provide more accurate and relevant responses to user queries. The pre-built libraries offer a standardized approach, making the development process more efficient. 

Providing application developers with a better, more streamlined building experience via a platform strategy is crucial to unlocking developer velocity and ultimately creating more end user value.

Unlocking what’s possible by developing new capabilities 

This wave of AI and GenAI innovation has given us a fresh perspective on what’s possible for our products. Ideas that seemed impossible during other points in our company history are now being reevaluated as practical and achievable. 

One of those ideas is our new conversation summarization capability, which takes long articles, posts, and comment threads and synthesizes them down to key insights. Application developers can now take this capability and use it across any of our chat-powered GAI products/features such as our AI-assisted learning experience, AI-powered premium insights, and more. Another example is how we’ve integrated our UCF (Universal Content Filtering) system into our GenAI platform to filter out low-quality results and keep our GenAI experiences safe, relevant and professional.

With new GenAI capabilities, our engineers can build new features for our members and customers that were previously incredibly difficult to build, or impossible altogether. 

Hiding operational complexity with the right tools

One of the challenges around implementing GenAI is optimizing for cost and latency. We need to be able to run GenAI products at scale but also remain cost-effective. These operational complexities shouldn’t burden developers or hinder their potential innovation, they should be abstracted away behind our platforms. 

One way we’re doing this is via our GenAI Gateway, which provides a common mid-tier for accessing and using cloud-based language models like GPT. It has features like rate-limiting for outbound requests to Azure’s OpenAI Services and a quota enforcement system, both designed to prevent individual product features from overusing computing resources or starving other features from access to those very same resources. 

By building out this operational functionality within our platform, we free application developers from the concerns of overseeing it directly.

Looking toward the future

While we continue to search for ways to make it easier for our teams to build, we’re also focusing on assessing the platform capabilities we’ll need to support our growth in the future.

Ultimately, remaining agile in the ever-changing tech industry is critical. We believe a thoughtful and responsible platform strategy is crucial to helping organizations innovate quickly to bring new AI-based products and features to market faster and with fewer frustrations. It’s enabling us to build smarter, deliver more consistent AI-based experiences, and power the next decade of innovation at LinkedIn. I look forward to sharing more about our journey and how we’re changing the way we work.