Track: Machine Learning for Developers

Location: Pacific LMNO

Day of week: Wednesday

Gone are the days when data scientists and PhD's were the only ones with Machine Learning skills. Today, ML is more approachable than ever thanks to a proliferation of frameworks designed to get you up and running quickly. However, we're still at the beginning of solutions that look beyond the algorithms themselves and support the full lifecycle of an application.

What are the components of a production ML application and which frameworks exist to support them? How are frameworks evolving to support this new paradigm of software development?

In this track, we explore the state of frameworks for ML and how they're being used in production today. Join us to hear how developers are deploying their solutions in sustainable ways and for inspiration on how to improve your applications with a dash of ML.

Topics include: ML primitives, Kubeflow, TensorFlow, GPUs, deployment architecture, CI/CD tooling, and more. No PhD required.

Track Host: Michelle Casbon

Senior Engineer @Google, Kubeflow Technical Advisory Council Member

Michelle is a Senior Engineer at Google where she focuses on open source machine learning tools and hybrid cloud deployments. Prior to joining Google, she was at several San Francisco-based startups as a Senior Engineer and Director of Data Science. Within these roles, she built and shipped machine learning products on distributed platforms using both AWS and GCP. Michelle’s development experience spans more than a decade and has primarily focused on multilingual natural language processing, system architecture and integration, and continuous delivery pipelines for machine learning applications. Michelle holds a masters degree from the University of Cambridge.

10:35am - 11:25am

Machine Learning 101

Today’s world generates different kinds of data at unbelievably rapid rates. This has resulted in a shift away from traditional software development towards fields like Artificial Intelligence and Machine Learning. Many are saying that Machine Learning is changing the world - but what does it mean? Why use it? What questions can it answer? This talk gives an overview of Machine Learning and delves deep into the pipeline used - right from fetching the data, the tools and frameworks used to creating models, gaining insights and telling a story. 

By the end of this session, audience members will have a better grasp of the capabilities and the commonly used approaches of Machine Learning. They will be familiar with different parts of the Machine Learning pipeline and will develop a strong foundation to continue learning and experimenting using tools like Jupyter notebooks.

Grishma Jena, Data Scientist @IBM

11:50am - 12:40pm

ML's Hidden Tasks: A Checklist for Developers When Building ML Systems

When we started building the NLP infrastructure for a startup around 4 years ago, few people had put deep neural networks into production. As the only machine learning oriented person, I was tasked with the entire pipeline - from gathering data, to training the model, to deploying it, to convincing our client to trust it, to maintaining and improving the model over time. As developers, we have a check list of things that need to happen in our minds, including building the software, various types of testing, infrastructure, data management, DevOps, and improvements. Turns out machine learning has an entire set of unexpected things that go on that "take it to production" checklist. This talk will bring to light my story of how I learnt about the unexpected are, and what tools helped us through that. 

Audience:

  • Software engineers who are moving into or currently working on machine learning systems
  • CTOs who are busy bringing ML functionality to their tools and products and want to understand what to look at

What you can expect to learn:

  • Unexpected details of putting models in production besides just the code, model and infrastructure: 
    • DataOps
    • Robustness and Uncertainty tests
    • Model Drift
    • Model testing approaches
    • Model Performance tracking
  • Specific tools and technologies that will help address the unexpected details 

Jade Abbott, Senior Machine Learning Engineer @teamretrorabbit

1:40pm - 2:30pm

From POC to Production in Minimal Time - Avoiding Pain in ML Projects

“So how soon can this go live?” It can be a chilling question because you know that whatever answer you give, there’ll be a business need to get delivered sooner and with fewer resources than you need.  Turning an AI proof of concept into a production ready, deployable system can be a world of pain, especially if different parts of the puzzle are fulfilled by different teams. When promised data doesn’t appear and timelines and scope creeps what can you do?

I’ll talk you through one such project: from the initial pitch and how that changed to the agreed project deliverables, the first AI model and a very clunky web demo, dealing with the extensive missing data and creating an automation pipeline to deal with it, getting a tensorflow based image classifier working in docker with a slick front end and continuously updating and deploying itself using codeship and AWS fargate. For each step, I’ll go into the technical detail so that whichever part of this puzzle you’re missing, you will be able to fill in the gaps and put something similar together yourself.

Key takeaways:

  • Setting up a data pipeline so that you can feed your models
  • Creating an api accessible ML model
  • Docker with GPUs and if you need it
  • Adding a demo suitable for clients not data scientists
  • Making production quality ML

How to put everything together and set up a continuous delivery pipeline for MLs models using docker, or staged deliveries using AWS fargate.

Janet Bastiman, Chief Science Officer @StoryStreamAI

2:55pm - 3:45pm

ML in the Browser: Interactive Experiences with Tensorflow.js

Machine learning (ML) holds opportunity to build better experiences right in the browser! Using libraries such as Tensorflow.js, we can better anticipate user actions, reliably identify sentiment or topics in text, or even enable gesture based interaction - all without sending the user’s data to any backend server. However, the process of building an ML model and converting it to a format that can be easily imported into a front-end web application can be unclear and challenging. 

In this talk, I provide a friendly introduction to machine learning, cover concrete steps on how front-end developers can create their own ML models and deploy them as part of web applications. To further illustrate this process, I will discuss my experience building Handtrack.js - a library for prototyping real time hand tracking interactions in the browser.  Handtrack.js is powered by an object detection neural network (MobilenetV2, SSD) and allows users to predict the location (bounding box) of human hands in an image, video or canvas html tag. 

Audience

  • Front end engineers interested in using ML within their web applications.
  • Software engineers interested in training ML models
  • Data Scientists interested in deploying ML 

What you can expect

  • A friendly introduction to ML in the browser using Tensorflow.js
  • When to use ML in the browser
  • How to create a machine learning model with an example (data collection, model training, model evaluation, conversion to Tensorflow.js).
  • Practical tips and pitfalls associated with ML projects (what model to use, data validation checks, what framework to use etc.)
  • A live demo of hand gesture interaction in the browser, using a neural network model.

Victor Dibia, Research Engineer in Machine Learning @cloudera

4:10pm - 5:00pm

CI/CD for Machine Learning

Machine Learning is now widely used across our industry, yet we have very limited tooling when it comes to automating the ML model versioning, testing, and release. We will show how a CI/CD pipeline for ML can greatly improve both your productivity and the reliability of your software.

Sasha Rosenbaum, Program Manager on the Azure DevOps Engineering Team @Microsoft

Tracks

Monday, 11 November

Tuesday, 12 November

Wednesday, 13 November