Inspiration

We as a team wanted to impact the most amount of people in a way that affects them everyday. Something that came to mind right was health and wellness space. As we started brainstorming we discussed an overlooked population. That population is our seniors (65+ years). Once we determined our target market we got into the nitty gritty diving into the benefits that are offered with Medicare Advantage plans. A benefit within the Medicare Advantage plan is a stipend to spend money on groceries. We then created Golden Groceries to help our senior citizens find healthy foods, track nutrient intake, and recommend them suggestions on recipes and restaurants for their nutritious needs.

What it does

Golden Groceries allows seniors to take pictures of the food they are eating to determine food nutrients. This tracks what nutrients the senior is intaking and then suggests foods that they could eat in order to achieve their daily goals. The tracking system is for daily consumption but also for the weekly consumption. This can also be shared with healthcare providers in accordance with HIPAA to help them better tailor to their healthcare needs. In addition to the creation of the recipes and suggestions there are options to purchase the foods our software suggests. Making a seamless experience for our seniors that allows them to use their benefits with our software. For example, a UnitedHealthcare member can use their U card in our software to purchase groceries, making sure our seniors utilize all the benefits their Medicare Advantage plans offer.

Our backend program uses FastAPI for API endpoints, where it receives and analyzes user-provided parameters. Initially, we calculate the user's basic consumption and nutritional needs based on their basic information. We then use ChatGPT to analyze the user's daily food intake by processing images they upload. When users inquire about nearby restaurants, we perform real-time searches using You.com and render the returned data in card format. Our system enables users to interact with it through conversations facilitated by an AWS-based Claude agent. This also allows us to include menus and other information in the conversation flow for real-time inquiries. The frontend is built using React to handle data flow and layout.

One big challenge we ran into was who to target with our software and who we will be helping and providing value. Really stressing who we will be impacting with our software was something we discussed in detail. Another portion that challenged us was developing and integrating ChatGPT with the You.com API. Since one of our team members was brand new to coding we were able to work through it and teach each other systems and applications.
We consulted with a mentor on how to call the Bedrock API interface for AWS and eventually succeeded in running the Claude 3 multimodal model on it. In collaboration with Intel Cloud Development, we successfully deployed LLAVA 1.6, a large language model for image and video processing with performance comparable to GPT-4. By modifying the model's source code, we managed to run it on an Intel 5th Gen CPU, which was previously only compatible with GPU CUDA, achieving speeds close to that of an A100 GPU. Our findings indicate that the native memory performance of the CPU is comparable to traditional GPUs. We integrated our entire search system using the You.com interface, allowing the model to autonomously determine the necessity of an internet search and return results automatically. In a video demonstration, we queried local restaurants to showcase this capability.

Our group is proud of all the learning that happened during the weekend and the growth on software systems and being resilient and persisting through challenging coding problems. A callout within our team is that we were not all on the same level of software development and if someone didn't know what another piece was, we asked questions and helped each other.

A piece of personal learning for our team was how to work with each other even though we were not all at the same knowledge level of coding. Overall our team has greater knowledge around software development which includes the You.com API, ChatGPT, React, Python, FastAPI, Node.js, and Claude.

The next steps for Golden Groceries is to conduct customer interviews and see how our target market interacts with the product. Once we get feedback from our target market we can use that to improve our value proposition and the overall experience for our seniors.