Scoping

1. Define the project, example: Speech recognition for voice search
2. Decide the key metrics:
   • Accuracy, latency(how long does it take to transcribe the speech), throughput (how many requests we will handle)
3. Estimate resources and timeline:
   • how much time
   • how much compute resources
   • how much budget
   • how long it will take to complete the project

Data

1. Define the data:
   • What data will be used?
   • Is the data labeled consistently?
   • Must do a pre-processing to collect and treat the data?

Modeling

3 key inputs that goes into training a machine learning models are:

• Code (algorithm/model)
• hyperparameters
• data

The result of this is a Ml Model

ML Systems = Code + Data

The error analysis can tell you how improve your data to have a better model performance. We don´t need always have more data to have a better accuracy of the model, but with the error analysis we can see what. data we can change to perform better.

Deployment

Deploy in production Monitor and maintain the system