1 of 17

Overview & H2/2021 Roadmap

Ville Tuulos

ville@outerbounds.co

2 of 17

The path to production is incremental and iterative

Prototype

Production

Debug

Cloud-based workstation

Explore with notebooks

Access data quickly

Create a workflow

Scale vertically

Scale horizontally

Scheduled execution

Freeze dependencies

A/B test

Version

everything

3 of 17

Business-critical project: Maximum SLA required.

Data scientist is prototyping a new idea locally. Failing is ok.

Promising project. Let’s scale it to all data.

Project is being A/B tested in production with a failover path.

Predictions feed into a decision-support system. Humans can error-correct if needed.

Production-readiness is a spectrum

Supporting projects at all phases of their lifecycle

4 of 17

Metrics

Tables

Text

Images

Addressing the common concerns in a single coherent framework boosts productivity.

Data scientist time is more valuable than machine time.

5 of 17

Regarding machines...

No need to reinvent the wheels

6 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

How much data scientist cares

How much infrastructure is needed

All layers of the stack matter

There’s a natural division of freedom & responsibilities

7 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Fast Data

Loading datasets quickly

8 of 17

Coming soon!

Coming soon!

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Compute

Executing batch jobs at scale

9 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Job Scheduler

Orchestrating DAGs reliably

Coming soon!

10 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Versioning

Keeping track of experiments, models, and data

Example

11 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Architecture

Developing non-trivial ML applications

A/B experiments

Example

12 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Model Operations

Observing and deploying ML applications

Coming soon!

13 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Model Operations

Observing and deploying ML applications

Metaflow Slack bot

now open-source! 🤖

14 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Model Operations

Observing and deploying ML applications

Example

15 of 17

Model Development

Feature Engineering

Model Operations

Architecture

Versioning

Job Scheduler

Compute Resources

Data Warehouse

Model Development & Feature Engineering

Training models at scale

Example

16 of 17

Curious to learn more?

Effective Data Science Infrastructure

How to make data scientists more productive

www.manning.com/books/effective-data-science-infrastructure

Book in progress

17 of 17

Thank you!

Join Metaflow community Slack at

http://slack.outerbounds.co