Inspiration
Hardware development is scattered across CAD programs, component databases, supplier websites, and prototyping tools. I built Helix to unite that process and make hardware creation more accessible, visual, and intelligent.
What it does
Helix is an AI hardware-development agent that turns a product idea into a build-ready design. It generates CAD, creates a structured bill of materials, previews the product at real-world scale in AR, and prepares its components for online ordering—all within one workspace.
How I built it
I used Jac as Helix’s orchestration and data layer. Projects, design versions, CAD artifacts, BOM components, conversations, and AR sessions are represented as connected nodes and edges. Jac walkers coordinate workflows such as processing a prompt, updating project artifacts, generating a BOM, saving versions, and preparing AR handoffs.
I connected Jac with AI generation, browser-based 3D rendering, device cameras, and component-sourcing tools. The interface includes dedicated workspaces for designing, reviewing the BOM, viewing the product in AR, and preparing parts for purchase.
Challenges I ran into
The largest challenge was connecting several complex systems into one reliable workflow. AR support differs across browsers and devices, so I needed fallbacks for devices without full WebXR support.
I also had to convert flexible AI responses into structured engineering data. A useful BOM needs consistent component names, quantities, references, prices, and suppliers—not just a paragraph generated by AI.
Learning Jac’s nodes, edges, and walkers while building alone under a hackathon deadline also forced me to prioritize aggressively and rethink the application’s architecture.
Accomplishments that I'm proud of
I created a complete workflow that moves from an idea to something a builder can understand, visualize, and prepare to construct. I’m especially proud of connecting CAD, AR, BOM generation, and component ordering instead of presenting them as separate demos.
I also used Jac meaningfully as the system connecting Helix’s projects, artifacts, and agent-driven operations.
What I learned
I learned that generating CAD is only one part of building hardware. A useful platform must also address scale, components, cost, availability, and the steps required to turn a design into a physical product.
Jac showed me how graph-based architecture can represent the relationships between users, projects, designs, components, and AI actions more naturally than disconnected endpoints.
What's next for Helix
Next, I plan to add live component availability, automatic substitutions, collaborative editing, improved world-tracked AR, electrical and mechanical validation, direct manufacturing quotes, and support for more advanced hardware assemblies.
My long-term goal is to make Helix the intelligent bridge between a hardware idea and a product someone can actually build.
