Problem area

Scientists lacked visibility of existing research

Computational models were spread across different platforms and teams, making it difficult to understand what already existed or whether it could be reused.

The challenge wasn't simply finding a model. It was making existing scientific work accessible and useful across the organisation.

Research findings

Finding a model was only the first step

Scientists needed more than confirmation that a model existed. They needed to understand what it did, who owned it and whether it was suitable for their research.

Without this information, discovering a model didn't necessarily mean they could confidently reuse it.

Design consideration

The platform needed to help scientists not only find models, but also understand and assess them before deciding whether to reuse them.

Strategic direction

How might we make existing scientific work easier to reuse?

Rather than replacing specialist tools, I focused on creating a shared experience that connected the information and workflows across them.

I structured the platform around four stages:

Discover → Submit → Review → Reuse

Three priorities guided the direction:

• Accelerate R&D: Help scientists find existing models instead of building from scratch.

• Enable reuse: Make models easier to understand and assess.

• Connect workflows: Bring teams together without replacing their specialist tools.

Design decisions

01 | Model discovery

Help scientists search by the problem, not the model name

Traditional search assumed scientists already knew which model they needed or where it was stored. But researchers often knew the problem they wanted to solve, not whether a relevant model already existed. I explored natural-language search, allowing scientists to describe a problem and use AI to find relevant models and studies.

Design decision

Let scientists describe what they're trying to solve rather than having to know a model's name.

Reasoning

Help researchers discover relevant work they might otherwise miss, reducing the need to create another model from scratch.

Human judgement

AI helped surface relevant models, but scientists remained in control and could inspect the evidence before deciding whether to reuse them.

02 | Model information

Give scientists enough information to make an informed decision

Finding a model didn't automatically mean it was suitable for reuse. Scientists needed to understand its purpose, ownership and status before deciding whether it could help their research.

Design decision

Make key model information visible as part of the discovery experience.

Reasoning

This helped scientists assess whether to reuse an existing model, adapt it or develop something new.

03 | Role-based workflows

Different roles needed different ways to manage the lifecycle

Scientists, submitters, reviewers and editors all used the platform, but each had different responsibilities. I designed role-specific views around the tasks each group needed to complete.

Scientists
Could find models and understand whether they were relevant.

Submitters
Could propose new models through a guided process and track progress.

Reviewers
Could review submissions, inspect evidence and provide feedback.

Design principle

One shared lifecycle, with clear views for each role.

04 | Connected workflows

Unify the lifecycle, not the tools

Replacing GSK's specialist platforms would have disrupted established ways of working. Instead, I focused on creating a shared layer that connected information and processes across existing tools.

Design decision

Create one shared experience for discovery, submission, review and reuse, while allowing teams to retain their specialist tools.

Reasoning

This connected previously fragmented workflows without forcing scientists to change their established ways of working.

Enterprise delivery

Build within GSK's existing design system

Introducing another component library would have added complexity and duplicated familiar interface patterns. I worked with developers to adapt GSK's Mendix design system to support the scientific workflows.

Design decision

Build on existing components and infrastructure.

Reasoning

With only four months to deliver the platform, this approach helped accelerate development and kept the focus on solving the scientific workflow challenges.

Impact

Accelerating R&D by making scientific knowledge easier to reuse

• 198 models across 34 platforms: Brought into one shared model management experience.

• Faster research and development: Helped scientists find and build on existing work rather than starting from scratch.

• Less duplicated effort: Made existing models easier to discover, assess and reuse.

• Enterprise MVP: Delivered within four months and validated through UAT.