STORY

Developed an agentic brain for autonomous search-and-rescue drones: a closed, end-to-end chain of LLM agents, each with its own task. It starts by reading the map and the camera views to spot clues. Each clue is then placed on the ground as a GPS position using trigonometry and given a discrete relevance level that updates the probability map. Finally, the next spot is chosen at the weighted centre of mass of the most promising area, and the loop repeats until the target is found.

TECHNOLOGIES

• Python
• Gemini & OpenAI APIs
• Google Earth Pro