Axiom

Links

• https://tryaxiom.co
• https://github.com/realTristan/axiom
• My Devpost account: https://devpost.com/realTristan
  
  

Inspiration

Building custom neural networks can be difficult, especially when working in teams. This is where my inspiration for Axiom originated. Currently working with a team of AI engineers, a constant headache is designing neural networks tailored to our needs. This is where Axiom comes in. Axiom is designed to make building neural networks fast, easy, and great for teams. I hoped to simplify Axiom to the point where literally anybody can come onto the website, and start planning & building.

What it Does

Axiom enables users to collaboratively architect neural networks, tailored to specific needs. It's implementation of Google AI's Gemini Pro model makes actually building these networks VERY easy. The human-like input, as well as the human-like output, makes understanding what's being built quick & easy.

Axiom leverages it's custom interpreter to handle the network sequence input. This is also used by the LLM to construct it's own interpretable sequence. Of course, I didn't want this to just be a Gemini wrapper, so I implemented a visual component to the application. Using a flowchart, automatically updated by what is generated by Gemini, you can visualize the layers of your network.

To make working in teams even easier, teams can document their network using a Rich Text Editor, which also supports Markdown.

Access your dashboard by creating an account with email + pass. Here you can see all of the projects you're apart of.

How We Built It

Axiom leverages a manifold of technologies.

1. Next.js (React)

2. Server Actions (Backend Logic)

3. Tailwind, NextUI, ShadCN, Framer-motion (Frontend Components)

4. MongoDB, Prisma, Docker (Database & ORM)

5. Vercel (Hosting)

6. Google Gemini Pro (AI)

7. NextAuth, JWT, SendGrid (Authentication & SMTP)

8. GoDaddy (Domain) (https://tryaxiom.co)

Challenges We Ran Into

The biggest challenges were:

1. Organization of components and state management.

2. Database logic.

3. Getting the flowchart to work properly.

All of which were luckily resolved, and fun to tackle :)

Accomplishments That We're Proud Of

1. The design of the website. It enables users to seamlessly engage with what Axiom offers. It's sleek, modern, & simple.

2. Simplicity. One of the main goals of Axiom is to allow almost anybody to start building neural networks without any pre-existing knowledge. The simplistic interface and intuitive content allows for this.

3. AI Integration. This was my first time integrating Google Gemini in any of my full-stack projects.

4. Other: Authentication system, SMTP, JWT.

What's Next?

There's a LOT more coming to Axiom.

1. More teams features.

2. Data management.

3. More dashboard features.

4. Greater customization of the network.

5. Better & more rich sequence parsing.

6. Creating massive projects.

And more!