Inspiration
Our project was inspired by Prajesh. Prajesh wanted to build a tool that can be used by first-time PC builders who were not sure where to look. He thought an AI chatbot that finds the best parts depending on the budget and the type of PC would be very useful and timesaving.
What it does
It takes requirements and constraints from the user and then researches the data the ai is specialized and trained on for the best combination of parts under the users budget. It then gives you details on the parts and some problems that could arise (For Eg. Bottlenecks). It also creates an image of the parts.
How we built it
We built this project by leveraging a multi-agent reasoning architecture powered by Google Cloud Vertex AI and LangChain. The system doesn't just retrieve data; it "thinks" through a three-stage pipeline where a Build Agent drafts an initial configuration, a Critique Agent independently identifies potential bottlenecks or price risks, and an Improve Agent refines the final selection based on that feedback. We integrated Vertex AI Search (RAG) to provide a deep knowledge base of historical component data, complemented by a Google Search tool for real-time market pricing and availability. Finally, we used Imagen 3.0 to generate high-fidelity, photorealistic renders of the final build, transforming technical specifications into a cinematic visual experience.
Challenges we ran into
We initially tried to scrape reddit for a dataset and train AI on the dataset to use for BuildBuddy AI. However, we had to remove this because there wasn’t enough data, so the AI couldn’t learn properly. Also it was difficult to find the latest products as many of the products were from years ago. We also did not have enough time to train the AI on a large dataset, where there wouldn’t be overfitting or underfitting.
Accomplishments that we're proud of
We’re proud that we were able to overcome our challenges and present a project that is truly helpful for anyone who is looking to buy a PC or PC parts for the first time and is not sure what are the best products to buy.
What we learned
We learned API integration, web development and AI agent fine tuning. We also learned better teamwork and how to split up responsibilities so that we are using our time and resources optimally.
What's next for BuildBuddy AI
Since this was the first time working on a project like this for all of us, we were unable to put out a well-made final product. We will try to get this website hosted online as it is currently only running locally. The UI/UX will also get changed up to match a more general theme instead of the hackathon inspired colorscheme.
Log in or sign up for Devpost to join the conversation.