Kaban

Inspiration

All of us team members are researches or have worked before in scientific research. While collecting or searching several papers we could find ourselves in issues of looking for a specific paper we read before and is hard to look for, just look for a specific section and not finding it quickly, not understanding some sections, not knowing which paper could be more convenient according to our objectives and other problems.

What it does

The tool works by obtaining one or several files and a user-inserted prompt which is used into said files. For instance, if you upload two papers and want to compare them by experimental results, the tool will hand out a generated results based on the input files.

How we built it

We constructed an LLM model which reads as input one or several files to understand their bodies and contents. Afterwards, by using the OpenAI API it uses a prompt to generate an answer according to the provided info. This is hosted and executed by a friendly UI to make the user experience much easier anc quicker.

Challenges we ran into

Some of the challenges we ran over were the logistics for frameworks integration and fine running the LLM model.

Accomplishments that we're proud of

We overcame every big in every one of the diverse frameworks we applied. We also developed the criteria to choose what features are factible to implement on time.

What we learned

Working as a team is always an experience, everyone has a lot to offer. We also got a better glimpse into the complexity and capabilities of LLMs, or furthermore Generative AI

What's next for Kaban

We expect for this project to grow much more further than only scientific papers and essays, we hope that eventually this is useful for several areas of expertise and questions regardless or scientific background.

It is also aspired that the model is much smarter in the future and can generate more complicated and useful prompt results for the user, like highlighting the results and a table comparing other tables.