Omdena — CropCycle
Django + FastAPI microservice architecture for ML crop-cycle prediction, with Redis task queuing and PostgreSQL.
CropCycle is an agricultural machine learning project predicting optimal planting and harvest windows from climate and soil data. The backend architecture follows the same microservice pattern I used in VisionVitals: a Django + DRF API as the primary interface, a FastAPI service hosting the PyTorch model, and Redis managing the task queue between them.
This separation means the ML environment (Python version, CUDA dependencies, model weights) can be versioned and deployed independently of the web application. Redis provides the job broker; Celery workers in the FastAPI service pick up inference tasks and write results back to PostgreSQL.
Built with an Omdena distributed team; my focus was the backend service integration layer.
Outcome
ML crop-cycle prediction model served via FastAPI microservice, integrated with a Django REST Framework API and Redis-backed task queue. Asynchronous job pattern allows long-running model inference without HTTP timeouts.