Project
AgriLumens
Personal R&D platform to process multispectral drone imagery and generate soil insights using modular ML pipelines and cloud-native services.
Founder & Engineer (Architecture + Backend + ML) Personal project Public summary
Personal R&D project. Implementation details are summarized at a high level.
Stack
FastAPIPythonGCPDockerNodeODMMachine LearningOpenDroneMap
Impact
- Automates the workflow from drone capture → processing → inference → structured outputs.
- Designed modular services to support multiple models and datasets over time.
- Built traceable pipelines with metadata tracking for reproducibility.
What I did
- Designed cloud-native architecture for imagery ingestion, processing, and inference.
- Integrated photogrammetry processing (ODM/NodeODM) into an automated backend workflow.
- Defined service boundaries to support scaling and long-running jobs.
Architecture
- Job creation → processing service (ODM/NodeODM) → inference service
- Artifacts stored in Cloud Storage with metadata tracking
- Optional event-driven orchestration for long-running tasks
Tags
MLGCPComputer VisionMicroservicesGeospatial