Chef Jarvis

Inspiration

My partner and I were inspired to create a nutrition app after we both figured out that we both had problems with keeping track of all the appropriate nutrient doses and types of exercises that our doctors recommended us, and thus we decided to make an easy to use, gamified nutrition - fitness app to help keep us fit.

What it does

Our app ingests data like your height, weight, daily diet, and age, in order to give our robot Chef Jarvis the context to give our users advice that enables them to lose weight, become stronger, faster, build stamina, etc.

How we built it

We built it mostly using the Anthropic tool Claude Code to write our code, using a tech stack we predetermined ourselves, using our past experiences in other programming projects. After Claude churned out an iteration of our software, we tested it and offered guidance to the agent so it could keep improving our project.

Challenges we ran into

We initially tried to use an API from the USDA in order to source nutrient data, but the USDA's food database's search capabilities were far too inaccurate for us to use, as it would often spit back incorrect data (e.g., user logs eating a banana, the software catalogs the user eating a banana pepper). Another challenge we had was one of our teammates being kicked out of the event due to his age, rendering us without a frontend developer, which significantly hurt our team, rendering us unable to construct our frontend without the help of AI.

Accomplishments that we're proud of

We are very proud of our ability to scale our coding capabilities, even with our lead front end developer gone, which built our perseverance. We are also very proud of our use of app's high degree of accuracy in tracking nutrients, with it having a very slim margin of error compared to real nutrition stats.

What we learned

We learned to read the instructions far more carefully to ensure we didn't unnecessarily lose another teammate again, and we learned lots of prompt engineering techniques in order to extract better code out of our robot partners.

What's next for Chef Jarvis

We want to expand Chef Jarvis to incorporate image recognition models like YOLO in the future, so that it can recognize food without the need of manual data entry, and for it to be able to connect to smart scales, eliminating the need for manual entry of the mass of our users. Using this, we could make our app much more useful than it currently is.

Built With

• axios
• bcrypt
• claude-code
• docker
• express.js
• grpc
• json
• nginx
• node.js
• openrouter
• postgresql
• react
• redis
• tailwind
• typescript
• vite
• zustand