Deliberate Machine Learning

Why subscribe?

Doing Machine Learning is frustrating. Our mission is to change that.

This blog is a collection of my thoughts on how we think about ML, how we do ML and what we want to achieve with ML.

Main topics:

• How to connect business value to Data Science?

• How to organise your team to deliver?

• How to write better code as a Data Scientist, and why does that matter?

• Article reviews

• More unpopular opinions

I want to focus on new and orthogonal ideas rather than reiterating cliches.

It is freely available and will remain so. Feel free to subscribe:

Who am I?

I am Laszlo. I am an ex-quant portfolio manager (commodities and futures), ex-mobile game analyst (Candy Crush & co.) and ex-startup Head of Data Science (market intelligence for Tier 1 investment banks).

I do machine learning professionally for more than 15 years.

I now run a startup consultancy helping companies get up to speed with their ML efforts. Get in touch if you want to learn more: https://hypergolic.co.uk/contact/

A sample of previous posts

How to write better code as a Data Scientist

1. How can a Data Scientist refactor Jupyter notebooks?

2. You only need 2 Design Patterns to improve the quality of your code

3. What is a Code Smell, and what can you do about it? (Part 1)

4. Refactoring for Data Scientists: How to maintain readability in a single method?

5. Simple trick to optimise code and maintain readability in a compute-heavy application

6. 5 Minimalist Tips for Data Scientists to reduce frustration while working with Pandas

7. Pydata London 2022 slides: "Clean Architecture: How to structure your ML projects to reduce technical debt"

8. How did I change my mind about dataclasses in ML projects?

9. Data Science code quality hierarchy of needs

10. Refactoring the Titanic - hands-on notebook refactoring exercise

Machine Learning Product Management

1. How to Connect Data Science to Business Value

2. How to solve Machine Learning problems for production? (Part 1)

3. How to solve Machine Learning problems for production? (Part 2)

4. You Only Need These 3 Data Roles in a Data-Driven Enterprise

5. The importance of a Data Acquisition Team

6. Separation of Concerns in Machine Learning

7. What is LeanML?

8. Scrum in Data Science

Article Reviews

1. Machine Learning: The High-Interest Credit Card of Technical Debt by Google

2. Hidden Technical Debt in Machine Learning Systems by Google

3. The ML Test Score: A Rubric for ML Production Readiness and Technical Debt Reduction by Google

4. Rules of Machine Learning: Best Practices for ML Engineering by Google

5. MLOps: Continuous delivery and automation pipelines in machine learning by Google

6. Machine Learning operations maturity model by Microsoft

7. Towards CRISP-ML(Q): A Machine Learning Process Model with Quality Assurance Methodology by Mercedes-Benz AG and TU Berlin

8. Rendezvous Architecture for Data Science in Production by Jan Teichmann

Causes of Machine Learning’s productivity problem series:

1. Strategy

2. Teams

3. Technology

4. Metrics

5. Process

6. Summary

Unpopular Opinions

1. Unpopular Opinion: Agile is not only suitable for Data Science projects, but it is the only way to run one

2. Unpopular Opinion: “Data-centric AI” is a straw man argument

3. Three paradoxes of rule-based Machine Learning systems

Featured elsewhere

1. MLOps vs DevOps

2. What’s wrong with MLOps?

3. Product Management in Machine Learning - MLOps Meetup #54

4. Machine Learning Product Manual - free ebook from Hypergolic

5. Data Science Risk Categorisation