EasyPC

Inspiration

Our inspiration all stems from one of our users terrible experiences building a PC. Building a PC can be an incredibly rewarding but also frustrating experience, especially for first-time builders. Our team member had a particularly disastrous PC build - incorrect parts that weren't compatible, unstable overclocking leading to crashes, and way over budget. It was a valuable learning experience but very costly in both time and money. We realized there had to be a better way than trial and error for PC novices to configure and buy the right components for their needs and budget. Our project aims to leverage AI and data to match users with PC parts tailored to their individual use case, whether that's high-end gaming, media editing, programming, etc.

What it does

The website takes in a user's custom PC build needs and budget, then gets to work finding the optimal parts combination. Users simply describe what they plan to use the computer for, such as high-end 3D editing under $2,000. Within seconds, the website outputs a complete part list with prices, specs, and descriptions tailored to the user's intended workload.

How we built it

Utilizing a Convex fullstack, we looked for cutting edge, pioneering softwares to help propel our project. We landed on utilizing together.ai to base our main AI system. Our fined-tuned llama-7b was trained on countless data-points, and works to create accurate and timely recommendations. Going further down, we used a secondary AI in MindsDB for bulk jobs to generate accurate descriptions of Pc Parts. We pulled from a scraped database in

Challenges we ran into

Along the way, there were many challenges. One included running on only caffeine, but that was deemed worth the trouble knowing the amazing project we built. On a more technical level, as the technologies we planned on using were newer, there wasnt that large of a network of integrations. To combat this, we produced our own implementations. Specifically for a MindsDB integration for together.ai. To further the usefulness of together.ai, we also created an integration for Discord. Furthermore, obtaining data was a monumental obstacle. As a group of 4 without much capital, we had to create our own legal web scrapping tools. We ran into countless issues but eventually created a capable web scrapping tool to gather publicly available data to train our model on. Eventually, we intend to invest into purchasing data from large PC Parts databases to finalize and secure data.

Accomplishments that we're proud of

We are definitely proud of the project we built, keeping in mind we are not all trained and seasoned hackathon veterans. More specifically, the revolutionary integrations are definitely a high-point for our project. Coming from knowing nothing about integrations, LLM creation and ethical data-scraping, we now know how to implement these systems in the future. And even when we would get frustrated, we always encouraged and pushed eachother forward in new and creative ways.

What we learned

We learned that even if we start from a basic understanding of how LLMs AIs and databases work, through passion and hard work we can become experts in this field.

What's next for EasyPC

Scaling this project will be easy. With an already fully-functioning AI system, the possibilites would be endless. We can keep feeding it more data, and we plan on implementing a text-feature where you could ask a fined-tuned LLM on any pc related questions.