It has now been exactly 4 years since we first open sourced Rasa NLU. Happy birthday to Rasa and our awesome community! 🎉
What an incredible ride this has been so far. We now have a huge global community of conversational AI developers and contributors, dozens of large enterprise customers, over 70 Rasa employees from over 30 different countries, and are backed by some of the best investors in the world. I’m proud of what we have achieved so far. But we’re just getting started - it’s a long journey ahead to accomplish our mission to empower all makers to build AI assistants that work for everyone.
It would feel odd to start a review of 2020 without acknowledging that this year has been extremely challenging across the world. We hope you and your families are safe and healthy.
Our developer community continues to grow fast, globally
We’re so excited to have welcomed so many new developers to the community this year. In 2020, downloads grew 5X to over 10 million, YouTube subscribers grew 6X to nearly 12,000, and the number of code and support contributors nearly doubled to more than 600. It makes us super happy to see that this growth is not limited to a few countries but is taking place across the globe.
Developers in our community work in all sorts of companies - large and small - on many different use cases. We added many new examples to our community showcase so make sure to check them out. A few highlights:
- Albert Heijn: Providing fast answers to customer questions of retail stores
- PicPay: Connecting Brazilian Families with Emergency Government Assistance
- Nora: Providing Education and Updates on COVID-19
As it has for everyone, the pandemic forced us to rethink our events plan for this year. Our first virtual conference ever was HUGE: The Level 3 AI Assistant Conference (L3-AI) welcomed more than 4,300 registrants from all around the world, with 50+ speakers and 40+ sessions. Big thanks to all our amazing speakers! You can find all the recorded talks here - some of the highlights:
- Designing Practical NLP Solutions (Ines Montani - Explosion AI/SpaCy)
- Fireside Chat: Rasa & JPMorgan Chase (Akela Drissner - Rasa & Liang Zhou - JPMC)
- Panel: How much should conversational AI developers know about ML and linguistics? (Emily Bender - University of Washington, Thomas Wolf - HuggingFace, Anna Rogers - University of Copenhagen, & Vladimir Vlasov - Rasa)
Towards better assistants with new product releases and education at scale
Earlier this year, we updated the 5 Levels of Conversational AI providing a framework to bridge today's reality with tomorrow’s potential. The key insight is that there is an inflection point between level 2 and 3, where developers and teams need to change their mindset of how to build and improve an assistant. To get to level 3 and beyond, you need to listen to and learn from what your end users are saying. We call this conversation-driven development (CDD).
Rasa Open Source 2.0 was our biggest release this year and but we shipped so many other exciting things:
- We introduced DIET, a new NLU architecture which jointly predicts intents and entities and lets you plug and play large language models like BERT and GPT-2. We’re excited that DIET has already been used to train models in 60+ languages
- The Rasa Playground, which lets developers get started with Rasa in the browser, before installing anything
- End-to-end training, a step towards the future of AI assistants without intents
- Incremental training, so that Rasa users can fine-tune their existing models instead of always training from scratch
- A radically easier way to deploy your assistant with Rasa Open Source and Rasa X, and many new features to help teams practice CDD with Rasa X
We wanted to give a big shout out for all the valuable code contributions this year, for example from Xiaoquan Kong, Pheel and Nicolas Beuchat. In 2020, over 100 people contributed code to Rasa for the first time. Thank you all so much!
A year ago, we promised more developer education, and we were thrilled to be able to provide education for every level of developer experience, across a broad range of topics. We can’t possibly list everything here, but I’ll try to highlight community favorites:
- Rasa Masterclass: With over 150,000 views, the Masterclass has been our most loved format to date. It gives an in-depth overview about Rasa.
- Rasa for Beginners: 3,300 students got their first introduction into Rasa through our new course.
- NLP for Developers: A series of short videos explaining key ideas in NLP, with practical advice on when to use various approaches, and common errors you might run into.
- Algorithm Whiteboard: Deep dives into the algorithmic details of how Rasa works, with lots of code samples and illustrations.
- Rasa Chats: A biweekly podcast with discussions about open source, applied research and how our community of makers is building accessible AI
- Rasa Certification Workshops: We launched different formats - live workshops and self-serve on Udemy. More recently, we also introduced advanced workshops about deployment and custom actions.
For more, make sure to check out our developer portal.
In addition to the content we create, we’re so grateful that contributors on all continents help other developers learn conversational AI and Rasa. For example, Cobus Greyling, Vishnu Priya and Ashish Saini. Just this year, we had over 100 new support contributors. Thank you all very much for sharing your knowledge and helping fellow developers.
Creating a lasting company
We strongly believe that creating a lasting company is key to maintaining the code we’ve already shipped and investing in new tools and education for everyone. This year, we made a lot of progress in that area as well. Our $26m Series B funding in June allows us to take a long-term view and invest even more into product and community growth. At the same time, we’re very stoked to work with a fast growing number of enterprise customers that want to bring their assistants to level 3 and beyond, and to deliver on the promise of conversational AI to their end users.
We grew our remote-first team to almost 70 “Rasas” from over 30 different countries, and will continue to hire smart folks across all functions. In March, we kicked off internal work at Rasa on diversity, equity, and inclusion. If our mission is to empower all makers to build AI assistants that work for everyone, we need to make sure this technology is inclusive and can be used by a diverse group of people. We believe that we have to start first with ourselves, but plan to share more with the community soon.
We will continue to invest heavily into product and community growth. You can count on us to ship more great product releases, create more education for developers and bring applied research into our open source framework. We will also make sure to keep up with enterprises’ rapidly growing demand to apply conversational AI across many different use cases.
Thank you for all the support this year! Go Rasa!
— Alan, Alex and the whole Rasa team