Leapfrog Technology, Inc.’s Post

Leapfrog Technology, Inc.’s first re:Invent was successful and eye-opening - and I want to share some of my personal takeaways. First and foremost, re:Invent shone an ultra-bright spotlight on (Gen) AI. From sessions to booths to partners, it seemed like there wasn’t a location in the building (any of them) without the words “AI” front and center. It was hard to have a random conversation without answering “What’s your AI strategy.” And that was both daunting and inspiring. Daunting since a lot of Leapfroggers build GenAI solutions on Amazon Web Services (AWS), and it turns out we’re very much not alone. And inspiring because there is so much AI innovation from AWS and its partners, and a clear demand for it. We’ve been experiencing that demand, and it feels like the tip of the iceberg. On the AI innovation side, here are the top five things I’m excited about: 1. New models on Bedrock  - Marketplace: Models tuned and scoped for specific use cases and constraints. At Leapfrog we work with frontier foundation models, and then tune them, prompt them, or RAG-with-them for vertical use cases. The prospect of using more customizable niche models like those from Upstage opens new doors that will ultimately allow us to build better solutions for our customers. - Nova: The new foundation model from AWS themselves that offers up to 75% cost savings of comparable class models. At scale, cost is something you need to watch out for, and we’ll be evaluating the performance/$ of this right away. 2. Tooling continues to evolve with things like SageMaker Unified Studio. At Leapfrog we use a combination of tools, IDEs, and notebooks to do GenAI/ML. The prospect of a completely integrated environment to explore, build, and test all things data/AI on AWS could be a huge productivity boost. 3. New features for Knowledge Base, like integrated document extraction, and features for Bedrock like prompt caching. We had to handle these things with code and context, and now perhaps, we won’t. 4. Using LLMs to evaluate…themselves. I’ve long loved the idea of using LLMs to automatically test and evaluate the outputs of LLMs. And AWS just released new tools for optimizing RAG and LLM implementations in this way. 5. The Industry and GenAI Pavillions - each a small city block in the expo. I observed GenAI tour de forces like having a robotic avatar interview Denise W. hold a real-time conversation, and draw a caricature based on her responses. Epic! And worth five dollars at your local arcade. In the Industry Pavillion, I was interviewed by Steve Faro as if he was my doctor, and watched AWS Transcribe + Bedrock document perfect SOAP notes based on my “encounter”. Leapfrog specializes in Healthcare solutions, and in particular data and AI-driven healthcare. The tools are here now to fix a lot of the efficiency issues for US Healthcare, and we are excited to be a part of it.