Designing a Platform for scientific research
TEMPUS
Bringing the power of data and artificial intelligence to healthcare
OVERVIEW
Tempus is a technology company building products that analyze extensive clinical and molecular data using genomic sequencing and machine learning. It’s vision is to use technology and data to revolutionize cancer care by providing personalized treatment recommendations for each individual patient.
My responsibilities:
-
Planning, executing, and synthesizing user research for supporting PM and stakeholders in making data-based product decisions.
-
Leading brainstorming sessions with PM and Eng to break down and solve user pain points.
-
Prototyping solutions and conducting user testing sessions to deliver a superior user experience.
Team and timeline:
-
Planning, executing, and synthesizing user research for supporting PM and stakeholders in making data-based product decisions.
-
Leading brainstorming sessions with PM and Eng to break down and solve user pain points.
-
Prototyping solutions and conducting user testing sessions to deliver a superior user experience.
Problem
Life science researchers at Tempus create various machine learning models for analysis of genomic and clinical data. These models can range from providing patient treatment insights to academic and research application.
How might we enable researchers to deploy their machine learning models on Tempus infrastructure to analyze genomic and clinical data at scale?
Solution
A guided digital platform experience that allows researchers to
-
Standardize and package their data and machine learning models according to Tempus architecture principles.
-
Deploy them in managed scalable environments on Tempus infrastructure.
Platform Concept
The platform is based on standardization of three core components of a machine learning model:
Data, Code, and Pipelines.
Platform design
A web-based app that simplifies the process by providing a step by step guided wizard with self serve tooling.
Structure your data with a visual schema builder
Easily browse available data at Tempus to use research and machine learning .
Guided 3 step process
Guided wizard to help you connect your own data to the Tempus data library.
Testing and validation tools to reduce errors and automate the process.
Easy management of your data products
Easily view and manage the data you are producing.
Integration with other Tempus system to allow for seamless experience.
Transforms- User Research
Need for research
Life-sciences creators want an easy way to develop & deploy their AI/ML algorithms scalably on the cloud using Tempus data. Tempus Eng built a service called Transforms to cater to this need, but there is a very low adoption and usage of it, and users are facing challenges in using it, to the extent that some are building their own ad-hoc solutions.
Research Goals
-
Identify and document Personas to better understand user groups.
-
Uncover user needs and pain points they face in creating Transforms.
-
Document their current user journey, and co-design a target user journey that better serves the target users.
Outcome
Consensus amongst PM and stakeholders on product priorities through data-backed decisions.
Increased understanding of users, user groups, their workflows and needs.
Common language and alignment across the team through documented artifacts and research data
Research highlights
3 User Personas
-
Categorizes the users into groups and help understand commonalities and differences across users.
-
Better understanding of group specific needs that help focus on most impactful areas of investment.
User journey map & pain-points
-
To help surface user pain points and areas of key challenges.
-
Identify user needs by user groups to allow product to make most impactful investments.
Accessible research documentation
-
To empower stakeholders and team in utilizing data in discussions and making informed decisions.
-
Facilitate meaningful discussions by developing common understanding.
In a team of 4 (2 designers, 1 PM, and 1 technical writer), conducted mixed method research using surveys & interviews, ethnographic studies, co-journey mapping, and affinity diagramming to produce:
Transformshub
Problem
Tempus wants to create a Marketplace of Transforms, which are reusable, plug-n-playable pieces of AI apps and algorithms in the bio-medical domain.
It wants it's internal developers and scientists to package their algorithms as Transforms. The current process of creating Transforms has a lot of challenges identified in the research above. (including ...but not limited to)
How might we design a service to make it easier for scientists (primary users) to create Transforms?
Key Challenges
1. Figuring out technical barriers to new process.
2. Simplifying guidance on highly technical concepts for users. / not exposing system complexity to user, platform that aligns with users mental model
3. Finding a balance between data security, ownership, and seamless user experience.
Impact
~80%
reduction in time savings
Improved experience with significantly reduced number of support tickets
Solution
Accessible data catalog
1. Figuring out technical barriers to new process.
2. Simplifying guidance on highly technical concepts for users. / not exposing system complexity to user, platform that aligns with users mental model
3. Finding a balance between data security, ownership, and seamless user experience.
Guided 3 step process
1. Figuring out technical barriers to new process.
2. Simplifying guidance on highly technical concepts for users. / not exposing system complexity to user, platform that aligns with users mental model
3. Finding a balance between data security, ownership, and seamless user experience.
Transform Marketplace
1. Figuring out technical barriers to new process.
2. Simplifying guidance on highly technical concepts for users. / not exposing system complexity to user, platform that aligns with users mental model
3. Finding a balance between data security, ownership, and seamless user experience.
