AI-Powered Portfolio

Overview

The goal is educational. I am trying to create a portfolio that is a self-hosted, AI-enhanced, and runs as a personal website. The idea is to showcase some of my experience, projects, and experiments.

It includes AI-powered interactions, a structured career timeline, a blog, and a project showcase.

This is still in development.

Goals & Scope

• Main Purpose: Showcase experience, talent and projects
• Key Features:
  • AI-powered chatbot that answers questions based on blog entries, projects, and timeline data.
  • Full CRUD support for blogs, projects, and career timeline entries.
  • AI embeddings for enhanced search capabilities.
  • Search combining full-text indexing and semantic embeddings.
  • Admin dashboard for managing all content.
  • UI for configuring the AI model and embeddings system.

User Roles & Access Control

• Admin: Full access to CRUD operations for blogs, projects, timeline events, and AI configurations.
• User: Read-only access to blogs, projects, and AI chat.
• Unauthenticated Users: Can interact with the AI chatbot and view public content.
• No role-based access beyond "Admin" and "User."
• No MFA support in this iteration.
• No OAuth-based access control for AI APIs – AI chatbot is publicly accessible.

AI Chatbot & Search Functionalities

• AI chatbot interacts without persisting chat history outside a session.
• Admin-configurable choice between Ollama and OpenAI API at the site level (toggle during setup).
• No user-submitted training data. AI training updates automatically when blog entries, projects, or timeline events are added.
• Hybrid search support: Combines full-text indexing and semantic embeddings.
• AI-powered search (e.g., "Show me projects using C# and MongoDB").
• Qdrant to index both project descriptions and full project source code.
• Everything locally hosted (no external embeddings).
• AI Chat:
  • Random greeting + guided prompt suggestions.
  • Freeform conversation somewhat allowed.
  • AI runs a pipeline: rail guard-> intent detection -> Context enhancement through RAG -> Response

Career Timeline & Skills Mapping

• Interactive career timeline UI for admins to enter career events.
• Events can be linked together (e.g., "Started at ABC" and "Stopped at ABC").
• Users can zoom, filter, and explore technologies used during different career phases.
• Graph-based skills map that visually connects skills, projects, and experiences.
• Career timeline should be D3-like interactive (zoom, forward/backward movement).
• Skills map should be drillable from broad categories to code snippets.
• Storytelling: Career journey timeline and anecdotes from key projects.
• Main sections of the page: "Time Machine" (Career Timeline).

Embeddings & Prompt Management

• Admin UI to manage embeddings.
• Admin UI to configure the AI model (Ollama or OpenAI).
• Embeddings generation and management handled internally (e.g., automatic updates on content changes).
• Ability to parse a local directory accessible from the backend.
• Support embedding source code, markdown, text files, and documents.
• Admins can add metadata to directories/files to enrich embeddings.
• Admins can modify AI prompts dynamically through the UI.
• Prompts should be stored in a structured format instead of hardcoded in the backend.

Admin Dashboard

• Initial version includes CRUD for blog posts, projects, career timeline, and embeddings management.
• Future enhancement: Analytics for most viewed projects, blog engagement, and AI chatbot interactions.

Content Prioritization

• Primary Audience:
  1. Fellow developers
  2. Industry leaders
  3. Recruiters / Potential Clients

Technical Considerations

• Frontend: Next.js with Tailwind CSS.
• Backend: Node.js with MongoDB.
• Hosting: Self-hosted on a local server.
• Search: Qdrant for indexing projects and source code.
• AI Chat: Ollama for chatbot interactions.
• Performance & Privacy: Everything runs locally; no external dependencies for embeddings or AI models.
• CI/CD Automation: GitHub, Jenkins, and Docker for automated deployment.
• Authentication: Cookie-based authentication instead of bearer tokens.

NOTE: This is all in progress, I manage the project elsewhere for now but plan to eventually do it here.