Introduction to Production Machine Learning
Session Overview
- AI, ML, Oh My
- Models
- What is Production Machine Learning?
- ML Ops
- Production ML
AI, ML, Oh My
![]()
Data Driven Science
![]()
Models
Encapsulated, compressed view of how the world works.
A function, with inputs and outputs
Models
“All models are wrong, some are useful.”
George Box
Sharing Models
Once you have a model, you can shore it by:
- Giving the model to other people to use
- Giving the results of the model to other people
- Giving the predictive ability to others to use
Trade offs between the three options: model complexity, ownership, accessibility.
Production Machine Learning
The rapidly evolving science of delivering data-driven insights.
ML Ops
The technical operations involved in delivering data-driven insights
- Building cloud platforms for model delivery
- Scaling
- Monitoring
- Course Correcting
Session Overview
- AI, ML, Oh My
- Models
- What is Production Machine Learning?
- ML Ops
- Production ML
Production ML
Once you have a model, you can shore it by:
- Giving the model to other people to use
- Giving the results of the model to other people
- Giving the predictive ability to others to use
Trade offs between the three options: model complexity, ownership, accessibility.
Production ML
Today, we are going to:
- Build an ML Model
- Functionalize that model
- Share that model using an API.
APIs
Application Programming Interface
A standardized way to communicate with a computer program
A ‘User Interface’, just not a graphical one
Today: REST APIs
REST API
![]()
REST API
![]()
REST API
![]()
Building REST APIs in Python
Hug
Production ML
Today, we:
- Built an ML Model
- Functionalized that model
- Shared that model using an API.
What are some tradeoffs of our approach for model sharing? How can this scale?
