Since I entered the blossoming public cloud space in 2012, I have seen many transformational technologies gain traction and scale by making build tools accessible to everyone. Marketing departments could hire their own developers and build what they needed in the cloud. Data warehouses were no longer the exclusive domain of specialized experts. Anyone could build an ecommerce site.
Infrastructure, platforms, middleware, frameworks, containers, software, functions and more are now widely available "-as-a-Service" because innovative vendors recognized that most end-users could not be experts in all the components of these architectures - and because managing the underlying plumbing rarely added significant value. This democratization of technology is driven by products and services that abstract away complex, evolving engineering layers and allow users to focus on business insight, creativity, and innovation.
Generative AI will follow this same trajectory as it matures. It has to so it can grow up.
The tools and landscape of GenAI are rapidly shifting sands, with new models, databases, architectures, and ideas emerging (and some fading) every week. As of this post, Hugging Face alone hosts almost a million open source LLMs.
Despite all the current noise around this model vs. that one, they are just one piece of the puzzle under the hood. While essential, they are design choices in a workload’s backend. As always, it is more effective to first define your desired business outcomes and only then figure out what design patterns, architectures, and components will best deliver those goals. And, those bits and pieces will surely change over time. The 'Model Madness' of 2023, which sparked so much excitement and experimentation, has since matured into more skillful approaches that can ensure high-value, scalable production workloads.
The Cambrian explosion of GenAI tools, building blocks, and design philosophies - and tempting use cases - is too overwhelming for most organizations to track. And, there is no end in sight to that churn as we all continue to sift through the hype to find the truly impactful ideas. This has been the story of AI for a long time. AI initiatives have too often been challenged in finding the right problems to solve and the best innovations to unleash - while using the right designs and data - and delivering measurable, compelling results. Plus, we all work for someone. It is crucial to make sure your early steps into GenAI are not a stumble.
Before building Treva, our founding team spent 18 months working with customers of all sizes to identify the right GenAI workloads and making sure those initiatives led to the most compelling business value. We infused everything we learned about our customers’ needs and wants into Treva’s DNA. We take care of the complexities, so you can focus on...
Outcomes. Not Algorithms.™
Scott
Treva Co-founder
