👨🔬 Multidisciplinary Engineer & Applied AI Developer
Welcome to my GitHub! I’m Robinson Messenger, an engineer with 8+ years of experience building real-world AI systems across agriculture, aquaculture, and logistics. I specialize in deep learning, reinforcement learning, and multimodal pipelines optimized for offline execution and field deployment.
🚀 What do I do?
- Design and implement end-to-end ML pipelines that combine satellite imagery, multispectral drone data, sensor arrays, and technical documents.
- Develop deep learning models (ViT, ResNet, EfficientNet) for classification, segmentation, and stress prediction in crops like maize, potatoes, and cherries.
- Use reinforcement learning and retrieval-augmented generation (RAG) to build adaptive recommendation engines tailored to agronomic contexts.
- Optimize models through quantization, pruning and tuning, enabling edge deployment without cloud dependency (TorchAO, ExecuTorch).
- Simulate supply chain scenarios using historical data and regression modeling to support decision-making from field to market.
🛠️ Technologies & Stack
Python · PyTorch · Lightning · scikit-learn · MLflow · Optuna
GDAL · rasterio · GeoPandas · FAISS · LangChain · ReportLab
Docker (local only) · JupyterLab · type hints · CI/CD scripts
🌱 Field Focus
All systems are validated and deployed on real farms in southern Chile, supporting disconnected, low-resource environments in Ñuble, Biobío, La Araucanía, and Los Lagos.
🎯 Contributions
Several geospatial and ML pipeline components are open-sourced to support the applied AI and agri-tech community.
