Machine Learning for DevOps and Applied AI — What We Learned in Houston Today
What do eBay, RBS, Lyft, Capital One and Adobe have in common? They’re all using AI and Machine Learning to transform business and disrupt their industries.
I mentioned in a previous post that RE•WORK is expanding quickly, with us bringing events to new locations and covering new topics. Last month we were in Toronto for the first time, and this week we’re in Houston for the Machine Learning for DevOps Summit and the Applied AI Summit. So why Houston? We always want to stay relevant to our audience, and with Houston being a manufacturing hub we were keen to host an application based event that showcased companies applying AI in the real world in areas such as e-commerce, manufacturing and retail, so Houston was an ideal location.
Today, we’ve heard from some of the most forward thinking and disruptive companies transforming their businesses with artificial intelligence. Here are some of our highlights:
Applied AI Summit:
- Hao Yi Ong, Research Scientist, Lyft
One of the rising stars of today spoke about his journey through academia into industry. Hao Yi Ong moved to Lyft around two and a half years ago to be part of a project that moves at a faster pace than the research world. At Lyft, the team are working to implement both machine learning and deep learning, and Hao Yi Ong explained that ‘a small team that constantly develops better features and maintains a good ML model retraining pipeline will beat any army of analysts that manually handcrafts and manages hundreds of business rules.’ We were also fortunate enough to sit down for a behind the scenes filmed interview which will be available on the Video Hub in a couple of weeks.
• Dancy Li, Data Science Manager, Facebook
Speaking about Facebook Marketplace, Dancy shared some really exciting progressions and updates that they’re currently working on. She explained that users see a ranked feed of products when they arrive at Marketplace — if Facebook knows what types of products each individual would be interested in ‘we can engage you in products on your homepage on Facebook that would make you more engaged with Marketplace’. Dancy also shared some of the exciting work they’re doing in personalizing notifications to drive each user to relevant items and potential purchases.
• Kyle Tate, Senior Data Science Lead, Shopify
Just before Kyle’s presentation, we sat down for an interview to learn about how Shopify are using ML in all areas of the business, but most specifically to bring ML powered products to a large production environment. Kyle spoke about how at Shopify ‘we believe this is sub-optimal, an understanding of the actual problems being solved leaks at each interface. Instead, we build data science teams that are responsible for all three of these tasks.’ We then caught Kyle’s presentation where he explained some of the ML applications: ‘We’re using ML to help in lowering the entry level into business. For example, with Shopify capital, you join Shopify, start building your business, and you find out how much money you’re eligible for then you accept and it’s straight in your account. We eliminate all the barriers that small businesses start in getting off the ground. This is built entirely by machine learning.’
Machine Learning for DevOps Summit:
- Koshsuke Kawaguchi, CTO CloudBees
Amongst all of our incredible speakers, we were honoured to be joined today by Koshsuke, the creator of Jenkins and CTO of CloudBees. He explained that this unique position means that ‘I get to see lots of real-world software development.’ ‘Helping people become more productive is my number one motivation. The automation in Jenkins is really simple. The reality is that automation and processes can get really complicated — the single person has no idea of the total extent of how complicated it is. Maybe you’re at the point where most of what we do it automated, but is it really achieving anything? We aim to simplify to boost productivity.’
- Nicolas Brousse, Director, Operations Engineering, Adobe
Adobe is using ML to prevent incidents, Nicholas explained that ‘Some of the findings that we have had, are that humans are not good at evaluating risk. Often they will bypass the risk and keep going’. Nicholas shared how the team leverage Adobe’s artificial intelligence and machine learning engine, to first, build predictive auto-scaling and self-healing services. He also shared how they’re using these technologies to provide insight and automate risk classification of production changes to reduce the impact on services availability.
- Chris Corriere, Senior DevOps Advocate, SJ Technologies
Chris kicked off by asking the audience who knew what the game ‘Telephone is’. There was a pretty decent show of hands that began his talk on The Worst Game of Telephone Ever. Before getting stuck into this case study, Chris backed up and explained that ‘your system is inherently broken if there are no people in your system because this means you can’t build a human-centric model.’ If we’re collecting data and there’s no feedback coming in, what’s the point? We need to learn from failure with a culture of continual learning and progression with humans at the centre.’ Chris gave an overarching definition explaining that ‘Devops is inclusion, complexity, empathy, culture, automation, learning, measurement and sharing.’
What else did you miss?
DevOps PANEL: From Theory to Practice: Making Machine Learning in DevOps a Reality
The discussion brought together ML experts and DevOps experts to explore views from both sides on how the two fields can work together to generate the most efficient model in DevOps. Diego Oppenheimer, Founder & CEO of Algorithmia was moderating the discussion and asked the room for a show of hands, ‘who considers themselves on the DevOps side, and who on the ML side?’ About half of the room raised their hands for each, demonstrating the enthusiasm for the topic from both sides. Panellist Adam Mcmurchie, DevOps Solutions Manager at RBS shared his thoughts:
‘In my mind, ML first means you have to utilise the data. If you occur data over time, do we really do anything with it? The pipeline is evolutional, so we need to use the data and run some neural nets on it to optimise the pipeline you already have. ML is just another strand that goes into the pipeline to make it better.’
Education Corner: Exploring STEM & Data Science in Houston
We really enjoy highlighting rising stars in AI, as well as encouraging the next generation to get involved. In the Education Corner attendees joined a group of local tech programs who were sharing their expertise in AI and ML. Shell shared their ‘AI Residency Programme’, a two-year immersive opportunity designed for students to work on projects across Shell’s business.
Interviews and Podcasts
Throughout the day, we recorded interviews with several speakers for our digital content platform, including our Women in AI Podcast. Margaret Mayer from Capital One joined us on the podcast and shed some light on supporting women and encouraging diversity in AI:
“I’ve had situations where I’ve been mistaken for an event organiser rather than a speaker, just because peoples biases take over. At this Summit it’s great to have a balance of genders both speaking and attending. It’s great to create a community where women can support each other, and also encouraging the women in tech to go to the conferences where they can feel that there’s a group of these women who all showed up!”
We’ll be back tomorrow to learn from more experts from Google, Walmart Labs, Spotify and loads more, so make sure you keep up to date with the days' events by following #reworkAI and #reworkDEVOPS on Twitter. Couldn’t make it to Houston? Check out our upcoming events, and save 25% on all future Summits when you register before tomorrow, November 30th, using the code CYBER25.