Project
Bird's AI – 5G Search & Rescue Drone
Bird's AI is a 5G-connected search-and-rescue UAV platform designed to assist emergency responders by autonomously detecting human presence in disaster-hit environments and streaming actionable intelligence with ultra-low latency.
View ProjectRecognition
📸Project Gallery
🎬Project Media
Nephele robots interacting with visitors at university open house
👤My Role & Contributions
Full-Stack Robotics Engineer
Key Responsibilities
- •Architected and deployed the entire AWS cloud infrastructure (IoT Core, Lambda, DynamoDB, Bedrock integration)
- •Developed the ROS2 onboard software stack for robot perception, navigation, and cloud communication
- •Integrated Amazon Bedrock LLM (Claude) for natural language understanding and context-aware responses
- •Built the React-based fleet management dashboard with real-time telemetry visualization
- •Implemented swarm coordination algorithms for task allocation and collision avoidance
- •Managed end-to-end deployment logistics for 3 public events with 6+ hour continuous operation
Technologies Used
🏗️System Architecture
Click to zoom
Problem
In large-scale disaster scenarios such as earthquakes, floods, and landslides, locating survivors quickly is critical to saving lives. Conventional search methods rely heavily on manual ground teams or line-of-sight aerial surveys, which are slow, manpower-intensive, and often constrained by poor visibility, damaged infrastructure, and unreliable communication links.
Approach
Bird's AI was designed as an end-to-end intelligent UAV system. A quadrotor platform equipped with a Pixhawk flight controller and Raspberry Pi onboard computer runs a lightweight computer vision model for real-time human detection. The UAV maintains continuous connectivity over a commercial 5G network, enabling low-latency transmission of telemetry, detection metadata, and visual evidence to a ground control station. Mission logic dynamically adapts flight paths based on live feedback to improve area coverage and detection efficiency.
Outcome
The prototype successfully demonstrated real-time human detection and stable telemetry streaming over a live 5G network. Compared to manual search patterns, Bird's AI improved area coverage efficiency and reduced response time, validating the feasibility of integrating edge AI with cellular-connected UAVs for time-critical emergency response operations.
