Create custom LLM Automation system

I am looking for someone who can develop a simple system that can do this. I’m not really knowledgeable in the space, but I am looking for someone who is. My assumption is that we need a combination of careful prompt engineering, automation and detailed customgpt development to accomplish this. Prompt engineering alone may not be able to achieve this. Goal: to develop a system that can invent new english words based on a topic provided which the user will input. These english words must have a real use case, it cannot just make up words that no one will ever use. 1. the user will enter a real english word and also a highly descriptive definiton of what that word is. 2. The system will then assume / guess potential concepts / topics around that word, broken down into 4 categories. For each category, we need roughyl 5 - 6 highly relevant words and/or phrases The 4 categories are: 1 - verbs around the word 2 - hobbies around the topic 3 - stereotypes around the topic 4 - metrics that could be used to evaluate something associated with the topic. 3. The system will then need to filter the bullshit defintions ones from the ones that actually make sense. For this step, it may be beneficial to get the LLM to intentionally hallucinate. 4. The system will need to generate a real word for each of the definitions. Each word must follow the rules of English. For example: - can be acronyms - Latin roots - conjunction words - verbs Etc…. Output Cat 1 (verbs): - pillow fighting: - Turning a pillow to the cooler side. -Cat 2 (hobbies): - pillow fort building - Pillow sculpting - Pillow optimization Cat 3 (sterotypes): - The 40yr old divorced dad who uses a 10 year old pillow still that has turned yellow - Pillow Collector - Owns dozens of decorative pillows but rarely uses them. - Pillow Hugger: Can't fall asleep without cuddling a pillow. Cat 4 (scores): - Sleep Support Score (SSS) Overall neck/head support quality - Compression Resistance (CR) Resistance to permanent flattening - Durability Index (DI)Expected usable lifespan Another important feature we need is this: Instead of providing an input word, we should be able to input a category, and in return, it will produce a few words and defintions in that category in any miscelaneous, totally random topic. This is what ChatGPT said based on my description above: https://chatgpt.com/s/t_6a6340373028819193e727a2cedb273e