Give each agent a responsibility
Specialist roles make intermediate outputs easier to inspect and connect.
AI systems2026
COMET.
Specialists, working together.
A multi-agent research and planning workspace coordinating specialist tasks, API integrations, evidence, and structured reports.
- MY CONTRIBUTION
- Agent orchestration & API workflows
- CONTEXT
- Collaborative project · Webcmd Hackathon
- STATUS
- Prototype
WORKFLOW / COMET
- 01Task + evidence
- 02Orchestrator
- 03Specialist agents
- 04Structured report
Specialists, working together.
01THE PROBLEM
What needed solving?
Research and planning cross tools and specialisms. A useful agent system coordinates those handoffs while retaining context.
02MY CONTRIBUTION
Where I came in.
I worked on an orchestration layer, API adapters, retry logic, and monitoring hooks. The workspace brings together business and student workflows, uploaded sources, and specialist chat.
Team repository03KEY DECISIONS
The thinking behind it.
Treat failure as part of the workflow
Retries and monitoring help identify where external calls or agent steps fail.
04OUTCOME & NEXT STEPS
Built. Learned.
Still improving.
An orchestration prototype demonstrated in a live pipeline, as documented in my résumé.
Where it stands today
A collaborative prototype. Agent workflows depend on configured model providers, supported tools, and source availability.
What comes next
Make recovery and the relationship between sources, agent outputs, and saved reports easier to follow.