A Machine Learning Platform by Intelus.ai
Project Overview: 
Duet is a no code natural language processing platform designed to be used by software engineers, data scientists, and data analysts. It enables users to create two types of machines learning (ML) models, classifiers and entity extractors. Duet is a highly technical platform that required understanding the fundamentals of machine learning as well as being able to balance user needs with the requirements from the companies executive leaders.
To learn more visit www.intelus.ai or try duet today at duet.intelus.ai/login
Team Role:
- Lead designer
- Design research
- Usability tests
Complex system
Duet uses machine teaching, which is complex and requires educating users on the practice.
Highly Technical
Machine learning is technical and can be hard to understand even for a technical audience.
Flexible Intergrations
Models created in Duet have to able to integrate with different 3rd party software.
The first steps I took to familiarize myself with the domain were Subject Matter Expert interviews and Contextual inquiry with our power users. Based on these conversation I created a user conceptual map and well as a technical functional map. My goal was to understand both how the system worked and how people perceived their actions.
Model Types in Duet
Duet allows users to create two types of machine learning models, classifier and extractors. Classifier models allow the user to categorize documents in to distinct entities and entity extractors allow the user to identify specific words or topics in the document.
When classifying document the user is presented with a document and a list of possible categories. The user then select the category that the document belongs to. As the model improves it will start making predictions for which category a document belongs to, this can greatly increase the speed that documents are labeled.
One of the differentiating features of Duet is the ability of the user to create Feature. Features allow the user to define what words, types of words, or phrases are relevant to a category. This only increases the accuracy of the model, but allows for each prediction to be traced back to the root.

The feature suggestion window surfaces the top features Duet is using, and prompts the user to either confirm that they are correct or remove them from the list.
Once the user accepts a feature they have the opportunity to edit it as need to most closely match the needs. This where the domain knowledge of the user comes in to play. The user knows what words in the dataset are sometimes used instead of "invoice" in this example and is able to add them, this way Duet will associate them together in the future and provide more accurate predictions. 
Entity Extractors
Designing Duet required a high level of technical understanding of Machine Teaching and machine learning in order to effectively build a system that is approachable by general users. To achieve this I worked directly with subject matter experts to understand the technical aspects of the system, while also working with users to understand their perceptions of how Duet works.

Annotated mockups of the labeling experience for Duet's Entity Extractor.

One of Duet's core functionalities for teaching models is user created features. Features are words, groups of words, or relationships between words that help the model create meaning in the data set. Below are the annotated designs for adding or editing a feature. While most of my designs for Duet avoid using modal, adding or editing a feature is a workflow that needs to be completed before continuing to teach the model.

Annotated mockups of the feature editor modal.

New User Onboarding Reserach
Part of my role at Intelus includes leading design research and usability studies. Often we work directly with current users to understand their needs and friction points. But we also use platforms such as userlytics to gain insights from new users who are not familiar with the platform. One of the studies was designed to understand how people learn to use the platform. This study led to a redesign of how information was delivered during the first time user experience.
Back to Top