I worked at Predixion Software as the sole product designer, reporting directly to the CTO. Predixion was a predictive analytics start-up with development office based in Seattle. There were mainly 3 products in the company: a cloud-based web platform for data scientists to publish, view, compare and visualize different predictive machine learning models; an Excel add-in for model creators to import data, create, and train the machine learning models before publishing to the web platform, and a web-based dashboard for industry uses (for example, nurses or maintenance crew can view the likelihood a patient will be readmitted or certain parts of a MRI machine will fail). All 3 products shipped and were available for all customers.

Fresh our of my design program, this was my first job in the UX industry. And as the only UX designer in the company, I was responsible for all thing design related: from visual, interaction and information architecture to marketing assets, the company website, pitch deck for raising funds, and weekly feature tutorial newsletters to our subscribed users. I also wrote and maintained all the CSS on the web apps. To deliver the best, most delightful user experience for data scientists, I spent a lot of time learning about how machine learning models work and how users would apply and compare these models. Looking back at this project now, I am grateful to have had the opportunity to experience all aspects of design, to learn about ML and the different phases of product development in an exciting, fast-paced start-up environment.

The web platform enables users, who are mostly data scientists and analysts, to make predictions on their data by comparing different ML models. For example, by learning a huge dataset of historical data of real estate sales, the model can predict how much any given future home will be sold for, learning from factors such as location, size, year built, etc. The platform offers different visualizations for different models, such sensitivity, neural network, forecasting models, etc.

Before publishing the models to the web platform to visualize, data scientists import data and create their models in Excel. Predixion also had a series of Excel add-ins and I was responsible for designing the entire set of over 30 icons. I also designed all of the dialog windows to assist users through the publishing process.

Predixion also provided specific solutions to enterprise clients. For healthcare, I designed a scalable dashboard view with 2 main pages; a landing page with list view, and a detailed page for each list item. For example, in the left picture below, the nurses can view all of the patients they are responsible for as a list view on the right, and see the likelihood they will be readmitted on the left. On the bottom left, it also shows what are the key driving factors that determine these predictions. Providing a solution like this enables more accurate planning and could potentially save operation costs for the hospital while reducing readmission rate. This dashboard was designed to accommodate various industries scenarios. For large equipment such as MRI machines and wind turbines, it's important to perform maintenance before they fail and not after. AI allows users to visualize how high risk certain parts of the MRI or wind turbine are for failure in the next few days. This equips the maintenance workers to be more confident and prepared when it comes to repairs.