What is your background in Machine Learning?
I started my journey in Machine Learning 10 years ago at university. I did an MSc in Intelligent Systems and continued with a PhD in Machine Learning at University College London. I’m now working as an independent consultant and I help companies make the most of ML. Over the last few years I wrote a book, Bootstrapping Machine Learning, that serves as a gentle and practical introduction to ML for innovators and hackers. I also started PAPIs.io, a series of conferences focused on real-world ML applications and on the tools and technology that power them.
What is your background in Machine Learning?
Both workshops (beginner and intermediate) aim at teaching how to integrate ML into your applications by adopting a practical, results-first and experimentation-driven approach. We’ll be using some of the hottest and best techniques, open source libraries and ML-as-a-Service platforms, including studios that make it faster to experiment with ML. (We'll use Docker to have all tools pre-installed in containers.)
Through example datasets and case studies, attendees will learn how to apply all of this to their own projects. At the end of the workshops, they'll be equipped with the tools and knowledge to start implementing and improving real ML systems, to launch experiments, interpret results, and also to know which questions to ask and problems to address.
The first workshop is for complete beginners and will start with an introduction to supervised learning techniques, to their possibilities and their limitations. We'll learn how to create, evaluate and operationalize predictive models. We'll also discuss how to best prepare the data to create these models from. (By the way, attendees can get a head start with the Machine Learning Starter Kit.)
The intermediate workshop follows on the beginner one with continued and additional topics such as text processing, parameter tuning, unsupervised learning, deep learning, and other techniques to get the most out of your ML workflows.
Why did you become a trainer?
ML can greatly improve technology by making it more intelligent, but education is currently the main bottleneck to integrate ML into the real world. There are amazing tools at our disposal for that (many of which are open source), and there is a lot of ML-related content on the internet. However, I keep meeting software developers and engineers who struggle to figure out how to apply all of this to their own projects, and statisticians and data scientists who struggle to deploy their work to production. At the same time, I'm seeing many gaps in the learning material available in books and online, which I'm trying to fill by teaching workshops.
I also just love sharing my experience and knowledge of the industry's best practices, techniques and tools gained from consulting and organizing conferences!
What are your plans for 2017?
At the beginning of the year I started working at my alma mater — UCL — as an Adjunct Teaching Fellow within the School of Management. We've been integrating some very exciting ML content into some of the courses, including the Machine Learning Canvas (a framework to better formalize usage of ML in real-world systems). This year I also plan to write a guide on using the MLC, and to collaborate on a report on APIs in Artificial Intelligence. Besides, I'll be co-organizing a few conferences — Sao Paulo in June, Boston in October (our annual PAPIs.io conference), and Paris (to be announced).
Where will ML take us this year?
We’ve been seeing more and more ML-powered features in high-profile consumer apps. We're also seeing commercial enterprise applications get more intelligent and predictive. This used to be a differentiation factor, but applications of all sorts will soon have to provide predictive features to stay relevant.
I'm expecting to also see an increasing number of in-house applications take advantage of ML (as an example, at PAPIs '16 last year Uber presented the custom MLaaS platform they built for internal use). Virtually all businesses and processes can benefit from the intelligence and automation provided by ML.
At the moment, it seems that most companies are still at the "proof of concept" phase of integrating ML into their domains. Deploying to production and considering long-term usage and impact will be key in the very near future.
Last year, we started hearing about ML-powered decisions that have the potential to change people's lives — rejection of a school / job / loan application, pricing of healthcare or home insurance, etc. — from a political standpoint, as the European Union passed a regulation giving its citizens a “right to an explanation” for decisions made by ML systems. It remains somewhat unclear what the form and content of explanations should be, but we can expect that privacy, fairness, accountability, interpretability and transparency will become increasingly important topics...
Photo credit: Brice Blanloeil