We (a small startup) have recently seen considerable success fine-tuning LLMs (primarily OpenAI models) to generate data explorations and reports based on user requests. We provide relevant details of data schema as input and expect the LLM to generate a response written in our custom domain-specific language, which we then convert into a UI exploration.
I'm curious if anyone has explored similar approaches in other domains or perhaps used entirely different techniques within a similar context. Additionally, are there ways we could potentially streamline our own pipeline?
While I like and use his work, I do think there are hundreds of thousands of programmers out there that are as productive and have as much impact, just that they don't happen to do open source.
It's not easy to list goals from the top of my head as it's not my strong side, and especially not if they have to be measurable (which good goals should be, rather than your boss evaluating whether or not you did something), but here goes:
Goals could be "Pick up language X in order to help development on project Y" or "Get formally introduced to all R&D team leaders, and get introduced to their roadmaps" or "Facilitate 10 job interviews together with team leaders in marketing"
Thank you. I asked because in the previous company I worked for we had difficulties setting measurable goals for devs and ended up tracking only product goals.
Product can be highly variable (or uncertain) so dev goals would focus on skills necessary to achieve said product. Ideally these skills a general and transferrable.
Have you ever read a white paper? You see how they manage to say that the product will solve every problem you have and nothing specific at all? Same idea.
I think to a large extent (i.e. enough to eliminate the worry in practice) it is how things work, actually. See my reply to the sister comment here: https://news.ycombinator.com/item?id=15040745
