Open to work

AI Assistant embedded in my portfolio

I integrated an AI assistant into my portfolio to answer, in real time, questions about my projects, contact, tech stack, experience and education.

Problem

Traditional portfolios are static and require users to manually navigate through projects, read documentation, and infer technical decisions on their own.

This creates friction, especially for recruiters or developers who want to quickly understand:

  • what was built
  • how it was built
  • and the thinking behind it

Solution

I built an AI assistant directly into my portfolio that enables users to explore my work through a conversational interface.

The assistant can answer questions about:

  • my projects
  • my experience
  • my tech stack
  • about me
  • etc

All in real time and based on my own content.

How It Works

  • The user accesses the /assistant route
  • They ask a question in natural language
  • The frontend sends the request using streaming
  • The model processes the query using contextual data from my projects
  • A real-time response is streamed back to the user

Tech Stack

  • Frontend: Next.js + Tailwind CSS
  • Backend: Next.js API Routes
  • CMS: Payload CMS
  • AI: Vercel AI SDK + Groq

Technical Decisions

  • Vercel AI SDK: simplifies streaming and real-time response handling
  • Groq as model provider: chosen for low latency and fast inference speed
  • Direct portfolio integration: improves UX and removes reliance on external tools
  • Context-based responses: ensures answers are relevant and grounded in my own work instead of being generic

Challenges & Learnings

  • Handling streaming responses on the frontend
  • Managing context to avoid generic or irrelevant answers
  • Balancing response speed and accuracy
  • Designing effective prompts for better outputs