A production-grade portfolio showcasing LLM systems, RAG pipelines, and evaluation frameworks.
Scaffolded with Lovable to accelerate boilerplate UI, then rigorously customized and architected from the ground up for maintainability.
This repository serves as a functional resume and portfolio for my work in the applied AI space. Rather than treating this as a simple weekend project, the codebase has been deliberately structured to mirror production software architecture—separating concerns, organizing by functional domains, and ensuring high customizability.
I prioritize systems thinking: understanding how things compose over simply knowing what tools to use.
The application diverges from flat, default AI scaffolding. It aggressively utilizes Feature-Based Architecture, ensuring components are grouped strictly by their domain relevance rather than technical type.
src/
├── core/
│ └── layout/ # Application shell (Navbar, Footer, NavLink)
├── features/
│ ├── ai-chat/ # Embedded Groq-powered AI Assistant
│ └── home/ # Landing page domains (Projects, TechStack, Hero)
└── components/
└── shared/ # Reusable primitives (ThemeToggle, ScrollReveal)
- Framework: React + Vite
- TypeScript: Strict type-safety across the client
- Styling: Tailwind CSS + Radix UI (
shadcn/ui) - Animation: Framer Motion (for specialized hardware-accelerated transitions)
- Backend Edge: Supabase Edge Functions (Deno) / Express
- AI Integration: Groq API (Llama-3-70b-versatile)
If you'd like to spin up the portfolio locally to examine the frontend or chat mechanisms:
# 1. Clone the repository
git clone https://github.com/ayushcodes13/devayush.git
cd devayush
# 2. Install dependencies (Bun or NPM)
npm install
# 3. Setup environment variables (Requires a Groq API Key)
echo "GROQ_API_KEY=your_key_here" > .env
# 4. Start the development server
npm run devThe application will be available simultaneously with the Express background server proxying AI requests properly at http://localhost:8080.