Put AI to Work
April 15-18, 2019
New York, NY
Please log in

Building AI assistants that scale using machine learning and open source tools

1:45pm5:15pm Tuesday, April 16, 2019
Implementing AI
Location: Sutton South
Secondary topics:  Deep Learning and Machine Learning tools, Text, Language, and Speech
Average rating: *****
(5.00, 2 ratings)

Who is this presentation for?

  • Machine learning engineers, data scientists, software engineers, and product managers



Prerequisite knowledge

  • A basic understanding of machine learning and its applications in natural language processing
  • Experience using Python

Materials or downloads needed in advance

  • A laptop with Python installed (version 3.6 preferred)
  • Clone the course GitHub repository (link TBD)

What you'll learn

  • Learn the machine learning models behind the NLU and dialogue management, techniques for training NLU models without pretrained word vectors, and best practices for gathering and preparing the training data and closing the feedback look using real-time user feedback
  • Understand how to implement conversational assistants from scratch using open source tools Rasa NLU and Rasa Core


AI assistants are getting a great deal of attention from the industry as well as the research. However, the majority of assistants built to this day are still developed using a state machine and a set of rules. That doesn’t scale in production.

Justina Petraityte offers a hands-on walk-through of developing intelligent AI assistants without any predefined rules, based entirely on machine learning and using only the open source tools Rasa NLU and Rasa Core. You’ll learn the fundamentals of conversational AI and best practices for developing AI assistants that scale and learn from real conversational data.


Stage 1: Natural language understanding
You’ll start with language understanding, bootstrapping from very little annotated training data. In this stage, you’ll also tackle the challenges of going beyond pretrained word vectors for NLU and enabling the assistant to capture more than one intention per user input.

Stage 2: Dialogue management
You’ll use machine learning to build the AI assistant’s ability to handle increasingly complex multiturn dialogues based on the actual conversational data. You’ll also enable the assistant to actually complete user-requested tasks by connecting the assistant to external APIs and using the knowledge of the outside world to steer the conversation.

Stage 3: Closing the feedback loop
In the last section, you’ll close the feedback loop by improving the performance of the assistant by using the interactive learning and a real-time user feedback.

Photo of Justina Petraityte

Justina Petraityte


Justina Petraityte is a developer advocate at Berlin-based startup Rasa, where she helps improve the developer experience in using open source software for conversational AI. Justina has a background in econometrics and data analytics, and her interests include chatbots, natural language processing, and open source. Her curiosity for data science and human-behavior analytics has taken her to many places and industries; over the past three years, she’s worked in the video gaming, fintech, and insurance industries.