A Look Back at Rasa Developer Summit 2019

Watch the recorded sessions from Rasa Developer Summit 2019.

On September 24, 2019 we held our annual Rasa Developer Summit in San Francisco, CA. The event brought together Rasa community members from across the globe for a one-day session filled with speakers from over 20 companies doing groundbreaking work in conversational AI. We’d like to say a huge thank you to all of our speakers and attendees who helped make the event a success.

But what made the Rasa Developer Summit so special? Let’s start with the numbers. We had a sold out event space, with over 250 registrations representing 150 companies. Of our attendees, 39% are actively developing a project using Rasa or already have a bot in production, and another 44% have used Rasa experimentally. With so many experienced Rasa users under one roof, it was an ideal opportunity to share knowledge, feedback, and community.

During two panel discussions and 14 talks, we heard from speakers at companies including N26, Adobe, Lemonade, and Facebook, who related experiences building custom integrations, shared cutting-edge research, and outlined strategies for leading effective product teams. We also heard from leaders at Rasa who shared a preview of what’s on the horizon for the product and community. In this post, we’ll recount some of the highlights from the Rasa Developer Summit--what we learned, what we’re excited about, and what’s coming next.

Our community is growing. Fast.

Growth was a recurring theme during the summit, and Head of Developer Relations Justina Petraitye had impressive stats to back it up during her presentation on the Rasa community. Between 2018 and 2019, we went from 1.1k to 6.7k community members, 150 to 300 contributors, and 100k to 1.5 million downloads. Amid all of that growth, we kicked off community chapters in NYC, Berlin, and San Francisco and contributors founded meetups in Brazil, China, and India.

The Contributor Program was a big hit over the past year, making it easier for community members to see where their contributions can have the greatest impact and lowering the barrier to entry for new contributors. We’ve noticed an uptick in new open source contributors as well as developers earning exclusive Rasa contributor swag for their efforts.

We also announced a new video series where developers can learn how to use Rasa to build sophisticated AI assistants — the Rasa Masterclass. Justina Petraityte leads the online workshop, which takes viewers through the steps to build their own Rasa-powered chatbot. We’ll be releasing a new video every Thursday, so be sure to subscribe to our YouTube channel!

Conversational interfaces are becoming the new standard.

Human-machine interactions are becoming more...human. Whether it’s an employee opening an internal IT helpdesk ticket, a store manager asking for yesterday’s sales total, or a customer filing an insurance claim, users prefer to interact with applications using natural language.

We saw examples of this from many of our speakers at the Rasa Developer Summit, including Adobe, who demonstrated a voice assistant for searching Adobe Stock. The project stemmed from the realization that many users weren’t opening the filter menu to search for an asset. Rather than re-designing the filter menu, Adobe took a different approach by creating an interface that users intuitively understand. Adobe’s voice assistant allows users to find photos matching their requirements using spoken commands, for example, “find me a photo of a man with a dog in a park with lots of green trees” or “show me pictures of the emotion: excitement.”

Praneeth Gubbala, of Sam’s Club at Walmart, demonstrated the “Ask Sam” assistant, which solves a different problem in a similar way. Club associates at Sam’s Club need thousands of answers every day to run their business, to questions like “When does Maria start her shift?” or “What price are avocados?” The “Ask Sam” assistant allows associates to ask questions using natural language, condensing a complicated web of resources into a single conversational interface.

Sophisticated contextual assistants are transforming customer interactions.

At the Rasa Developer Summit, we heard from several speakers running sophisticated contextual assistants in production — many of them fully capable of handling unexpected inputs and incorporating the user’s prior statements into ongoing conversations..

In a talk by N26, Edouard Malet explained that a chatbot was a logical solution to the problem of scaling support across 3.5 million banking customers in 26 countries. Customers can resolve requests like downloading a bank statement or requesting a replacement card through an assistant available 24 hours a day—and they can do it in 5 supported languages. In solving this problem, N26 also tackled the organizational complexity of managing workflows across data science and content teams. The N26 team’s talk described how they developed a wrapper that allows content creators to create custom actions in markdown--no Python coding required.

Lemonade is an insurance industry disruptor, and their approach to automating processes across the company using AI assistants is no different. Like N26, Lemonade uses customer service chatbots to provide high-availability assistance to a rapidly growing customer base. During the Summit, Nathaniel Kohn of Lemonade demonstrated how the team uses sophisticated AI assistants to help customers accomplish tasks like filing an insurance claim, with minimal friction and in record time.

Advances in deep learning and NLU are redefining what’s possible.

The Summit Research track featured data scientists and academic researchers who led us through their work on the frontiers of conversational AI.

Bing Liu of Facebook shared recent research on the role of deep learning and reinforcement learning in training task-based dialog systems (think Siri or Alexa), and Nouha Dziri, PhD candidate at the University of Alberta, outlined a state-of-the-art framework for evaluating dialog system coherence using entailment techniques. We also heard from Mady Mantha, AI Platform Leader at Sirius Computer Solutions, who demonstrated how to produce a custom NLU pipeline integrated with BERT.

There’s more to come with Rasa.

One of our favorite parts of the Rasa Developer Summit is sharing what’s new with the community, face to face. Head of Engineering Tom Bocklisch presented a first look at recently developed features as well as a preview of what’s ahead on Rasa’s roadmap.

What’s clear is that it’s been an eventful year for the Rasa engineering team! Since launching Rasa 1.0 and Rasa X in May, there have been 2,761 commits made by 122 contributors. As of Rasa 1.3 and Rasa X 0.21, we’ve made it easier to integrate domain-specific knowledge into assistants, with actions for querying Knowledge Bases. We’ve also added policies that better handle chitchat and FAQs and laid the foundation for scalable deployments. In future releases of Rasa X, we’re planning to improve support for continuous delivery workflows by integrating Git for model versioning.

Conclusion

There were many more presentations, panel discussions, and hallway-track conversations than we can catalogue here, but we’d like to extend a huge thank you to all of our speakers who contributed their stories and experience. And to the community—we’re honored and humbled to know that so many of you are using Rasa to build amazing things and contributing your energy and expertise back into our open source projects.

If you weren’t able to attend this year’s Rasa Developer Summit, you can find more photos and commentary by following the #RasaDevSummit hashtag on Twitter. Slides are already up for many of the talks, and we’ll be releasing video in coming weeks. Until next year, you can keep up with fellow Rasa community members at one of our global meetup groups or events—we’ll see you there!