MNIST (as toy CV project)

Purposes:

1. Allow new comer / AIEngr team to pickup E2E ML (Training) in AIP
2. Enable validation and demonstration of more MLOps processes both in development and production (e.g. Model Experimenting, Model Serving, Brittleness, Contextualise Testing, Drift Detection, etc) with toy project that involve different ML frameworks and domains (NLP, ASR, DA, CV, maybe autonomy)
3. Create scripts, templates to help as a whole AI community to

• speed up adoption of MLOps with minimising re-learn and mistakes.
• standardise works to streamline process or minimise requalification of works.

4. Demonstrate values of various MLOps processes, including deployment strategies.

Code Structure (Live diagram - to be updated every commit)

E2E flow

https://docs.google.com/presentation/d/1AS7s-hTxcK2ec1ULZ1K_yubbBXipRdWQ2Gk6a46SJQE/edit#slide=id.g10bb7c7143d_0_50

python docstring

• Adopted numpy-style.
• There are 3 available styles https://queirozf.com/entries/python-docstrings-reference-examples#numpy-style.