Overview
At Insporos, I contributed to the development of an AI-driven pipeline designed to automate seed quality control for greenhouses. The system focuses on predicting seed germination and identifying potential diseases early in the agricultural cycle, significantly reducing the operational costs and waste associated with planting defective seeds.
Technical Contributions
Gantry System Design & Data Acquisition
I designed and developed a custom gantry machine specialized for high-precision seed handling and data collection.
- Sorting Mechanism: Engineered a system capable of detecting, picking, and sorting individual seeds into microwells without damaging the biological samples.
- Data Acquisition: Integrated hardware for spectrographic readings and high-resolution image analysis, providing the necessary raw data for the AI germination models.
- Prototyping: Handled the mechanical assembly and sensor integration required to ensure the machine could operate reliably in a high-throughput greenhouse environment.
Computer Vision & Path Optimization
A core part of my role involved moving the system from a proof-of-concept to a production-ready speed by optimizing the seed collection logic.
- Computer Vision: Implemented state-of-the-art models using OpenCV to accurately identify seed orientation and location in real-time.
- Efficiency Gains: Developed and implemented shortest-path algorithms to optimize the gantry’s movement during seed collection.
- Results: These optimizations reduced the processing latency from 5 minutes per seed to approximately 20 seconds per seed, a 15x increase in operational throughput.
Key Skills & Tools
- Software: Python, OpenCV, Shortest-Path Algorithms, AI Pipeline Integration.
- Hardware: Gantry Robotics, Spectrographic Sensors, Microwell Sorting Systems.
- Optimization: Latency reduction, high-throughput automation, motion planning.