Loading...
All Projects
AI & AutomationPRIVATE SYSTEM
AURORA
A self-orchestrating fleet of twelve AI agents running Northstar's daily operations.
A multi-agent orchestration system I built from the ground up. 12 specialized agents working in concert—each optimized for a specific domain. Personal assistant, legal aid, research, analytics, code review. Learns and evolves continuously. This is what I use to run Northstar.
- Client
- Northstar Coding (internal)
- Role
- Creator & Architect
PERSONAL PROJECT
- 12
- Specialized agents
- 24/7
- Autonomous operation
- 5+
- Domains covered
The Challenge
One person running a software firm wears a dozen hats: research, legal, analytics, code review, scheduling. Off-the-shelf assistants handle one conversation at a time and forget everything between sessions — none of them can operate as a coordinated team.
The Solution
AURORA is a multi-agent orchestration layer built from the ground up: a central conductor that decomposes incoming work and routes it to one of twelve specialized agents, each tuned for a single domain. Agents share context through a common memory layer and hand results back for synthesis, so the system behaves like a staff rather than a chatbot.
The Outcome
AURORA runs Northstar's operations every day. It drafts research, pre-screens legal questions, reviews code, and tracks analytics continuously — compounding in capability as its learning loop tightens. It is the proof-of-work behind every AI system Northstar ships for clients.
Highlights
- Central orchestrator routes work to whichever agent is best suited for the task
- Agents cover distinct domains: personal assistant, legal aid, research, analytics, and code review
- Continuous learning loop — the system refines its own routing and prompts over time
- Runs the day-to-day operations of Northstar Coding in production
Technology
LangChainNext.jsPythonOpenAI
Have a project like this in mind?
START YOUR PROJECT