Dispatch-AI

Inspiration

-Critical incidents are decided in the first few minutes, yet response systems rely on delayed human recognition, overloaded monitoring, and manual calling. Most emergencies are already over before help arrives. We built Tracer to remove those early delays by automating detection, contextual reasoning, and response coordination so action starts when it still matters, therefore saving lives through the power of saving time.

What it does

-Continuously monitors CCTV feeds -Uses computer vision to detect high-risk events (weapon, fire, person down, suspicious activity) -Extracts contextual evidence from video clips (what happened, where, confidence) -Scores threat severity in real time -Delegates tasks to the correct response agents (police, fire, medic) -Presents a structured incident summary to a dispatcher for final review and action

How we built it

CCTV feeds stream into an Overshoot CV layer for real-time visual detection Detected events are passed to Tracer’s anomaly detection layer for severity scoring and context extraction A central agentic orchestrator evaluates the evidence and determines the response type Specialized agents (police, fire, medical) are activated based on the incident context All outputs are sent to a dispatcher interface for confirmation and coordination with first responders

Challenges we ran into

Connecting the backend to the frontend. UI Design

Accomplishments that we're proud of

Integrated computer vision with agent-based decision orchestration Built an end-to-end detection → reasoning → dispatch pipeline

What we learned

Human-in-the-loop design is essential in high-stakes systems. The end desicion should alwasy be done by a human, the Agentic Software just aims to streamline and make that process more efficient.

What's next for AI Agentic Surveillance

Real-world pilot deployments (schools, workplaces, campuses) Expanded CV capabilities and multi-camera correlation Smarter confidence scoring and incident prioritization Quality of Life Updates Integration with existing emergency infrastructure

Built With

• arize
• backend
• frontend
• overshoot
• python
• thetokencompany
• typescript