As an ML Developer, I'm interested in training from scratch or fine-tuning state-of-the-art deep architectures based on TensorFlow, PyTorch, or MXnet for different Vision, Speech, or NLP tasks.
But due to my 5 years of experience of real-world challenges, As an MLOPs Developer, I'm in love of
Quantize, Prune, and Distill my trained model to lighter and faster production-ready microservices based on TVM, TFlite, ONNX, OpenVino, and CoreML to bring huge models to server-side or client-side applications.