LLM tools

Building assistants

LLM tools

Here are a couple of projects I’m running in the LLM space.

GBN Stakeholders - Feb 24

So, we had this neat idea not too long ago. I thought, why not make a little helper that could dream up personas in green building neighborhoods, or GBNs as the folks in the know call them. I wanted to make something that could take a sprinkle of imagination and turn it into a full-blown persona, complete with likes, dislikes, and maybe a set of motivations, all wrapped up in a tidy template.

It sounded a bit like planting a garden in the digital world, where each persona would bloom from a seed of an idea. It’s been a good DIY challenge, and let me tell you, this was quite the adventure.

GBN Use cases - Feb 24

I recently embarked on a little project, something a bit like teaching a computer to write stories. My aim was simple: to create a helpful tool that could take a basic idea about another tool and turn it into a well-thought-out document, the kind businesses use to understand if an idea is worth pursuing.

Imagine telling a colleague about a use case you thought up, and then, voila, you have a neat summary of why it’s a great idea. It’s still in its infancy, but has the merit of producing something people can build on and critic. It’s getting there, learning bit by bit to make sense of our big ideas. It’s been an interesting piece, filled with coffee and head-scratching moments, but seeing it come to life has been interesting.

LLM calls caching framework - Jan 24

Not too long ago, I decided to tackle a new project, something a bit like building a digital bridge - a gateway that lets people chat with those smart computer programs, the ones that can write stories or answer questions, known as LLMs (Large Language Models). But here’s the catch: I wanted to make sure it was like a smart bridge, one that remembers the travelers so that not everyone has to pay a toll every time, making the journey more affordable.

It’s still a simple setup, with a few (not so clever) tricks to make sure it doesn’t forget who’s passed by before. It’s still a work in progress, learning new ways to be more efficient and helpful.