Autonomous Disaster Response

The winning feature: a visible multi-agent rescue workflow

Input Agent: Accepts disaster reports, locations, and uploaded images.
Situation Analysis Agent: Extracts reported damage and urgent needs, clearly distinguishing verified facts from assumptions.
Priority Agent: Scores simulated incidents using transparent rules, such as urgency, number of people affected, and available resources.
Resource Planning Agent: Uses a tool to allocate simulated rescue supplies and suggest response priorities.
Verification Agent: Checks the plan for missing information, resource conflicts, and unsupported claims before presenting it for human approval. Build only these essential features first:
A dashboard showing simulated disaster incidents on a map.
An image and report upload feature.
Three or four agents with visible task logs.
A priority list and resource allocation plan.
An evidence panel showing why each incident was prioritized.
A human approval step before any consequential action. Use simulated data and clearly label the application as a decision-support prototype, not a replacement for emergency services. Your final demonstration Present a simulated flood scenario with several affected locations and limited rescue supplies. Show the agents analyze reports, prioritize needs, identify a resource conflict, revise the plan, and produce a final response report. Compare the plan with a simple baseline to demonstrate measurable improvement.

The key differentiator: Make the agents' actual tool calls, decisions, validation checks, and results visible on screen. That demonstrates real agentic behavior rather than simply wrapping an LLM in a chatbot.

No project can guarantee a win, but this is my single recommendation for you because it combines AIML skills, a compelling real-world problem, and a demonstrable multi-agent workflow within 24 hours.