JARVIS AI Assistant
A fully custom personal AI assistant built in Python. Designed to be genuinely useful rather than just technically impressive. Dual-model routing, persistent memory, browser automation and live data access.
What was built and why
Most AI tools available today are generic. They answer questions well but they do not know who you are, what you are working on or what happened last week. JARVIS was built to fix that by creating a personal assistant that actually has context and memory.
The project started as a way to automate repetitive tasks and has grown into a fully functional assistant that can handle complex workflows, access live data and remember past conversations.
How it works
- Dual-model routing: GPT-4o for reasoning, Gemini for speed and cost efficiency
- Long-term memory using ChromaDB vector database for context retention
- FastAPI backend serving a clean HTML dashboard interface
- Browser automation for web research and data extraction
- Live market data integrations for real-time financial information
- Modular architecture that allows new tools to be added without rewriting
What this project demonstrates
Building something like this requires understanding how to structure backend logic, manage API calls efficiently, work with vector databases and create a frontend that makes the tool actually usable.
The goal was always function over form. Every feature added had to justify its complexity by making the assistant more capable in a practical, everyday sense.
Interested in AI tools or automation?
Reach out if you want to discuss a project or collaboration.