© Krista Mellin. Automatic segmentation of patients.

Design for machine learning & artificial intelligence

While technology sprints forward, oftentimes human-computer interaction lags behind. Novel data-intensive business can get a huge boost from finding better ways to gather, store, improve and finally benefit from the abundance of data around us. We just need to make sure the end product is human-friendly, ethically sound and provides real value.

Understanding the magical ingredients

Designing for ML and AI is a bit backwards in the eyes of a traditional UX designer, as the new technological possibilities are hardly visible to the non-data scientist. The best place for a designer is between academic curiosity for what is possible and a team of very good data experts.

Mapping possibilities and actual needs

Sometimes it is easy to get carried away by experimenting with everything new, but 99 percent of the challenge (apart from actually engineering the data) is finding the right use cases. I've been bridging the user needs, business requirements and technological solutions into tangible concepts and service blueprints. This has helped in evaluating, explaining and testing new ideas before development has really begun.

© Krista Mellin. Natural language processing and labeling.
© Krista Mellin. Data explorer for medical data.

Distributed computing: A case study.

I had the opportunity to redesign a user interface for administering distributed computing. Cloud computing is used for a multitude of use cases, for example academic research and machine learning. In this project the end goal was to make the interface for managing the computing capacity easier for novice users and more modern for experienced ones.

Communication with niche experts

Being an expert in a very niche area is enviable, but deep expertise may blur the lines of what kind of concepts others are able to grasp. Designer has the opportunity to be the translator of complexity to the mere audience we others are. I spent a lot of time learning the key components by talking to the experts, all the while trying to keep a memory of what was hard to understand at first. By simplifying and clarifying we were able to make the workflows more streamlined while offering instructions to the user in the tricky parts.

From text to structures with natural language processing: A case study.

Natural language processing (NLP) is an interesting tool for surfacing insight from freeform text, such as patient journals. A lot of information is hidden only because it is buried in a sea of text. Using NLP we can find key concepts from the piles of old text and provide measurable insight for, in this case, doctors. Finding the actual need and how to approach the problem was a cross-disciplinary effort from business, technology, design and healthcare domain professionals.

My responsibilities:

  • User interviews
  • Co-creation & facilitation
  • Concept design
  • Visual design
  • UI design & design system
  • Interactive prototypes (Figma)
© Krista Mellin. Distributed computing UI.
© Krista Mellin. Distributed computing UI.