Inspiration
We were in the Draper University Fundamentals of Entrepreneurship program, and this was created during the Herothon, a 3 day-Hackathon which was part of the program. The theme of the Herothon was to create an app for ChatGPT.
What it does
We are developing a market validation app to help startup founders quickly do idea validation with an AI customer reasoning engine.
How we built it
Since it was a short hackathon, we just used prompt engineering.
Challenges we ran into
While the initial results were impressive, there was no clear roadmap on how we could evolve/improve our platform
Accomplishments that we're proud of
We managed to execute and win 1st place at the Herothon
What we learned
We had a crash course on NLP and learned a lot about LLM-Ops. We also learned the value of having good quality data, which we produced with a combination of manually created datasets. open-sourced datasets, and also some synthetic data. Also, having few-shot learning also significantly increase the reliability of the LLM response.
What's next for NicoleMind
We hope to be able able to have the opportunity to fine-tune the MPT-30B model, which we suspect will be even more performant.
Built With
- databricks
- huggingface
- mlflow
- mpt-7b
- python
Log in or sign up for Devpost to join the conversation.