I’d like to seed my LLM with my personal career experience so that potential interviewers could ask questions that would produce answers and build the LLM data as those conversations continue…ie:
Q: do you have an experience as a .net architect?
A: yes, i was a texhincal architect for 2 years with company abc from startdat to enddate. (Need more info/etc?)
Q: do you have experience mentoring developers?
A: no experience found (ask for clarification)
This could be a RAG style app that maintains the necessary metadata to tweak the responses/prompts to/from the LLM.
(apologies for misused terms and if this is a dupe!)
The idea being that you could turn your own experience into a living document that grows and evolves, and questions and answers dont need to be repeated endlessly in interview scenarios.
RAG will certainly make light work of this project. However, a looming question I have for you is; is this supposed to be a fun project, or a legit product? Is this for practicing interviews, or are you trying to replace them? Because using this as an aid or more for a fun project is one thing, but trying to make an AI act as your proxy for conducting interviews is another. There’s a lot that could go wrong, even if you build it right.
I found chatbase which is essentially what i was looking for but yes it is just a fun project, i am a .net developer looking for work and i am sick of getting the same questions all the time and wanted to just set up a spot for a pkm of mine while i am “managing my personal knowledge” during the course of my current retooling phase.
Id like to develop a react based mobile app using azure openai or the openai rest api to add this feature as a way to build the kb for my career.
As in, currently my resume lacks experience working with OpenAI and React, soooo……as i develop this, i develop the skills, and those skills add to the source (corpus?) of the LLM (right?)
And of course tagging text as embedded elements to create facets that can be used for comparison etc… (i.e react mobile skills)
RAG aspects?
Preserve and track the corpus of the llm as it evolves and seed/weight/tweak as needed to refine OpenAI calls.
Annnnd building…and we start to blur lines between mining archived knowledge and managing current initiatives - the scope of such a pkm could be fraught with scope creep…but whats scope creep to a curious developer?
So i was using chatbase but it is limited to like 20 api calls per month or something….i am hoping i can (1) find a tool that does essentially what chatbase does for free or (2) build my own chatbot style app with my own api - but how many call will openAI itself limit me to per month if i go that route?
Via chatgpt: (more than 20)
As of my last update, OpenAI’s API pricing might have changed, but historically, OpenAI offered a free tier that provided 1,000 API requests per month for both OpenAI’s GPT (ChatGPT) and OpenAI’s Codex (formerly known as OpenAI’s API). However, it’s recommended to check the current pricing and offerings on OpenAI’s website for the most accurate information.
So while chatbase is a great little tool - too much $$ for a dev playing around!
So i was thinking…what if you could change this business model on it’s head?
Let’s say that you had an assistant that housed a set of assistant (each one of them like you)…then the reviewer could target you as a reviewee and ask some elementary question…going through cheaper turbo3.5 for limited questions…then paying an advanced version of you (gpt4) for some more details.
Then you could charge others like you for the questions that that reviewer asked you and the answers that you provided?