BuildGuild

Landing page
Discover feature
Community feature
Showcase feature

Inspiration

A 2020 McKinsey report revealed that companies with diverse teams are 36% more likely to outperform their peers. Additionally, studies have shown that the failure rate of product development projects decreases by 30% when team collaboration is high. However, discovering the right people to collaborate on a project can be daunting. Inspired by these findings, we are motivated to build a platform that facilitates connections within communities, enabling better collaboration on projects.

What it does

BuildGuild is ChatGPT for finding individuals with whom you can collaborate. With its chatbot capabilities, you can input a prompt however you want and find individuals exactly how you want. After discovering your guild, you can communicate with them in our community chat room. There, guilds will build their connection and ideate a project. Lastly, user can showcase their projects. This showcasing is done every week to track the progress of the project.

How we built it

We heavily take inspiration from ChatGPT's design, so we didn't heavily brainstorm for a new UI/UX design. After having the rough design in mind, we divided into two teams: full-stack and ai. The full-stack team creates the website and integrates it with our BaaS (Supabase) and AI API. Meanwhile, the AI team creates a Retrieval-Augmented Generation (RAG) that automatically takes data from our PostgreSQL database and generates a response based on that data. The website is built using Next.js and the AI is built with FastAPI and Ollama.

Challenges we ran into

At first, we weren't certain our ideas were going to work. We even had to build another idea before this (in total ~7 hours wasted). After reflecting that our other idea didn't belong to any of the hackathon's team we decided to commit to BuildGuild. The AI Engineer on our team hasn't even touched building a LLM let alone RAG before, so a majority of time is spent researching and learning how to do exactly that. After 20-ish hours we finally had a proof of concept that the AI works, it was then that we decided to fully commit and speedran the project (as hackathons do).

Accomplishments that we're proud of

We are extremely proud to see our project completed. Although minimalistic in color, the design of the website still looks sleek and modern. Besides that, we are also proud of creating an RAG LLM that is capable of finding users based on a simple prompt.

What we learned

We learned that we probably should have a concrete idea of a product before committing to it, as we previously mentioned, we have wasted 7 hours on a dummy idea. Besides that, having a proof of concept or MVP of your core business logic can be nice so that much time isn't wasted on researching and learning.

What's next for BuildGuild

So far, BuildGuild is still using local resources since we didn't have the budget to afford a cloud server with a dedicated GPU. In the future, we would like to deploy our website on the web so that our solution can truly be used by communities.