Key takeaways

This was a challenging project as the stakeholders were all a specialist in their own domain - our users who were the domain experts, the data scientists who articulated the logic behind the machine learning pipeline, and the developers who built the platform - and my role as a UX designer was to visually translate each member’s needs, requests, and concerns to ensure that everyone was in agreement before taking a step forward. This ultimately resulted in a series of iterations from understanding the current machine learning pipeline flow, to designing the ideal flow, to ideating the platform UI, to testing the developed platform.

In retrospect, there was a lot we could’ve done differently. However, this project had become a turning point in my career where I realized the importance of democratizing technology - making technology and design accessible to people with limited knowledge.

Here are some of the biggest takeaways from the project:

Visualize complexity

Establishing team alignment was challenging, especially when the team members were all a specialists. I learned that the best way to communicate in this type of situation was to avoid an oral meeting and instead, focus on visualizing everything that was spoken, whether in a form of a table, flow chart or prototype. The key was to externalize our thoughts so that we were using the same reference to make important decisions.

Be intentional about who to invite to a meeting

Although making the process open and inclusive is extremely important, having too many people in a meeting can delay the decision-making and create unnecessary confusion. For major decision-makings, I invited just the leads of each domain to a meeting, ensuring that they communicate closely with their members to represent their body appropriately. The leads were also invited to a client meeting (previously, only business members directly interacted with the client) to answer technical questions on the spot. This helped reduce the number of meetings and clarify ambiguity in a timely manner.

Size down the research scope

For this project, we had numerous questions that had to be answered by our user before moving into development. Were the flows for each step intuitive and easy to understand? Did they understand how to recover from error? Were there too much or too little explanation? etc. I found that narrowing down the research scope and being explicit/intentional about what’s not included in the research helped us navigate complexity efficiently. Also for prototyping a machine learning platform, I have found the Wizard of Oz experiment effective. You can still gain lots of insights by role-playing the autonomous parts!