Deploying your own ML systems

Hi there! For the second year now, I’m sending a letter to my Medium audience to share significant projects I’ve been working on. (The previous one was about my book on the Machine Learning Canvas.)

Anyone looking to innovate with Machine Learning and to apply it in their own domain has had to think about how to align ML’s possibilities with their Objectives and Key Results. If you’re in a managerial position in tech (e.g. as a CTO, product manager, or lead developer) you’re probably aiming at:

• Leveraging ML to create value for your clients/users/teams
• Taking control of your company’s ML transformation and building its ML assets (data acquisition, preparation, modeling and deployment pipelines — a.k.a. workflows)
• Leading and executing end-to-end ML projects with confidence: choosing the right problem to tackle, the right tech, avoiding pitfalls in your ML system design, prioritizing ideas, saving time, and deploying predictive models with confidence.

How? — Don’t focus on algorithms, but on the full ML stack!

90% of the code in an ML system isn’t ML code per se, but it’s informed by the application of ML. If you’re spending your time in lockdown getting up to speed on topics you care about, here are some learning objectives to add to your list. You want to understand how to…

• Pick your own pilot project based on feasibility and impact
• Frame your own ML problem and design a system that creates value for its end-users
• Collect your own training data and prepare it for ML usage
• Build your own baseline ML models quickly
• Assess whether you can trust a model for usage in production
• Industrialize your models and deploy updates continuously in a production-ready system.

Real-world ML that delivers value

You might come across blog articles that cover some of these topics, but it’s difficult to find hands-on and practical content. And it’s even harder to find learning material on real-world ML (as opposed to ML in a controlled experimentation environment, e.g. on fixed datasets).

I haven’t found any single resource that covers all of the above (i.e. the “full stack”) and that’s accessible to tech managers… So I created my own course!

My objective is to help you own Machine Learning: the course empowers you to design and to deploy your own ML systems that create real value for your company. You’ll save time spent on execution, and money spent on consulting.

If you’re interested in learning more, check out the course webpage for more information (syllabus, what’s included, screenshots, FAQ)… Doors will close on Sunday 26 April 2020!

Stay safe,

Louis