Aide-n

Inspiration

In emergency situations, quick access to accurate information can be critical. A web app that accepts both text and image inputs allows for intuitive interaction, particularly in stressful situations where typing detailed descriptions might be challenging. This could be especially beneficial for users who are visually impaired, have limited mobility, or are not native speakers of the language. This innovative approach leverages technology to make medical information more dynamic, accessible, and practical, bridging the gap between laypersons and the often complex field of medical care.

What it does

Aide-n is a platform that allows users to input an image of an injury which will then be taken and transformed into a text description. The description is presented as an output, and also taken as an input into an Intel Large Language Model. The LLM uses text to text modeling to generate a potential solution to the injury that was described.

How we built it

The model was built using two parts. The first part invovled creating an image to text generation utilizing Google's Gemini API. The second part involved taking the textual description of the image to generate a potential solution to the injury. To do this we built an LLM using Intel's architecture including their Intel Developer Cloud as well as pytorch-gpu.

Challenges we ran into

We ran into many challenges. The first challenge was figuring out how to utilize Intel's architecture including their jupyter notebooks to run our code. Since our code was a LLM, it had to take a significant amount of power which required us to run on Intel's developer cloud rather than our local machines that we could integrate with our UI/UX. This lead to many issues afterwards including the integration of Gemini and our LLMs. We had also incorporated our own instances of Small VM - Intel Xeon 4th Gen Scalable processor which were extremely hard to navigate due to their remote nature.

Accomplishments that we're proud of

We are proud of being able to learn so much about creating and executing Large Language Models. In addition the idea that we could connect onto a remote processor that had a significnat more power than our local machines was really cool! We're also proud of being able to pull all-nighters :)

What we learned

We learned about LLMs, Gemini, remote processors, and so many companies as well, all in a really short time.

What's next for Aide-n

Aide-n is essentially encompassing two models, that are trained on datasets that we can change at our will. This means that Aide-n is not just limited to First Aid but rather can be trainined on medical images, research papers, and so much more. The potential use cases of these models has so many possibilities and it's exciting to think about them!