Intelligent-search with natural language processing for the VO

Hello everyone, I would like your help with an idea I have. You see, for my thesis project, I want to develop a chatGPT-like web app oriented towards astronomy and astronomic data found in the VO. The idea is to use Intelligent-search and natural language processing to access the data from the Virtual Observatory in a way that emulates the human-like behaviour and conversation that chatGPT and other Generative AI’s produce. I would also like to give the AI access to PyVO documentation or train it with code examples, so it can consult the VO through TAP, SIAP, SCS and other protocols.
I’ve only scrached the surface so far in terms of GPT, AI and NNs. Would any of you have some advice or ideas on how to carry on this project? Maybe some literature on the matter or specific models of NN.
Thanks to all who took time to read this question!

Bito, I would encourage you to go on the OpenAI chatGPT website (and if you are a paid subscriber) you should be able to search through all publicly available GPTs. Search specifically for one called ChatGaia. This GPT constructs ADQL queries that are wrapped in a URL that when sent by you will download astronomy .csv data (in answer to your query). It will also enable you to perform EDA and plots on your data. I encourage you to continue in this topic as I consider it pretty much SOA and beats the heck out of generating your queries from scratch. As there are more than 128 different astro oriented TAP services out there - you have plenty of opportunity to expand on this. Many fellow astro-types are very interested in this. Also consider using it with scriptable TAP interfaces like the scriptable version of TOPCAT (STILTS) using a GPT.
Keep up the good work!