Local & Custom AI Coder
A self-built AI coding platform designed in Figma and engineered as a scalable, model-agnostic development system.
Student Project
Role
Designer & Developer
Industry
UI/UX Design
Project Time-line
1 month
Overview
A custom AI workspace that simulates a multi-role development pipeline built with AI and designed in Figma. This project explores backend concepts like guardrails, pipelines, fallbacks, and failure logging, translating these technical concepts into an intuitive UX. The project offers a way to more deeply understand how local AI systems behave in practice, and how to develop a UI that make complex systems usable for non-technical users.
Problem
AI coding tools lack transparency, control, and scalability across different hardware and model capabilities.
Contraints
Running on consumer hardware imposed strict limitations on model size, performance, and system design.
Goals
Develop a flexible “Codex-like” framework where output quality scales with available models and hardware without changing the core system.
Outcome
A modular AI framework that automatically swap between models while keeping the same underlying system, allowing it to scale as better hardware becomes available. Deployed through a Cloudflare tunnel with internet access, enabling real-world usage and testing. This project explores system architecture & UX, and required continuous AI testing to evaluate their different capabilities, strengths and weaknesses
