Designing the virtual lab platform to speed up adoption by neuroscience research
Role:
- product-designuser-researchproduct-ownership
- frontend-engineeringweb-design
2022–2026
Switzerland
Summary
Turning brain models that were hard to read and to connect into a single platform to explore, simulate and validate them, adopted by more than 1,200 labs. I worked as the only product designer, with a product owner’s responsibilities, as the Blue Brain Project became the Open Brain Institute and the team went from 142 to 32 people. I redesigned the circuit viewer, its display modes and the provenance map, and its bounce rate fell from 86% to 8%.
Ask me
- Company
Open Brain Institute (spun out of the Blue Brain Project, EPFL)
- Year
2022 – 2026
- Role
Sole product designer, product owner of the Virtual Lab, Help & Feedback and Reports, and frontend engineer
- Team
32 people (142 at the start): 18 scientists, 4 full-stack, 4 backend, 3 in visualization, 2 in HPC
- Star metric
Circuit viewer bounce rate from 86% to 8%
- Problem
An imprecise 3D brain with no connectivity, large circuits that were unreadable, provenance limited to a single level. The result: an 86% bounce rate on the circuit viewer, with users taking screenshots and contacting support.
- Decision
Replace the 3D with a layer-by-layer 2D view, add display modes that each answer one question at a time, and an interactive provenance map.
- Result
Bounce rate from 86% to 8%, and more than 1,200 virtual labs in 18 months against a target of 600 a year.
01
Context & business friction
A platform to launch from scratch, with a small team and a survival target
In 2022, EPFL’s Blue Brain Project became the Open Brain Institute, and the team went from 142 to 32 people. The public launch took place in March 2025, with no labs and no partners. To survive, the institute had to gather 600 virtual labs a year. Yet the circuit viewer put people off: an imprecise 3D brain, unreadable circuits, an 86% bounce rate. I worked on it as the only designer, with a product owner’s responsibilities.
02
The Key Trade-off
What I pushed through, negotiated or refused
I refused to add guests to virtual labs. A lab stays a space for administration and budget, and members join projects: opening access to guests would have blurred those responsibilities. I also said no to complex case studies on the home page. Trade-offs were made every week with the scientific leads and the CTO, in a Friday ticket defense, using RICE and impact/effort as the grid.
03
The Transformation
From brain models that were hard to read to a guided exploration, one question at a time
The viewer now opens in 2D. A selector lets you go back to 3D, but it is the slice that guides the reading: you move slice by slice (“Slice 12”), and each region, labeled and colored, reads one at a time.
A “View” menu changes how the same circuit reads. With “Synapse concentration”, the 2D slice becomes a colored 3D view: each mode answers one question at a time, instead of showing everything at once.
The provenance map shows where an entity comes from and what follows from it: three parents, twelve siblings and 3,211 children. A slider sets the level shown, and you move from one sibling to the next without leaving the map.
The platform opens on a home with three entries: Data, to explore morphologies, electrophysiology recordings, electrical models and synaptic properties; Workflows, to launch the building of ion channels, single neurons, synaptomes, microcircuits and brain circuits; and Notebooks, to run and adapt neuroscience notebooks. The OBI Assistant sits below them, and on the right edge of every screen. In Data, each model opens in a panel with its morphology and trace, its region, species, M-type, E-type and license. In Workflows, a configuration is a list of steps with their status, beside a 3D view of the neuron being built.
04
Business Impact & Proof
The measurable value generated for the institute
The circuit viewer no longer drives users away: its bounce rate fell from 86% to 8% (Matomo). The platform also beat the institute’s survival target, with more than 1,200 virtual labs created in 18 months against 600 expected a year, and 12 partner universities.
The data validator took imported entities from 2,100 to more than 30,000, and a biological data company partnered with the institute to enrich the 2D views. Researchers thanked us for the default 2D, their requests now concern new features such as dendrograms, and usability is no longer raised in workshops.
05
Critical hindsight & retrospective
What I learned and how mature I am
Today, I would build each workflow’s onboarding backwards from the outcome: a conversation with the OBI assistant about a real publication, then the way back to what produced its results and artifacts.
In the notebooks, I also proposed built-in tips that point to the platform’s native tools. I would have pushed them earlier, because notebooks are a way into an environment that researchers already know.
Spotlights
- 01
Building a new UX for success
Scientists left within 8 seconds. We scrapped Explore, Build and Simulate to start over.
Read the spotlight → - 02
Single neuron simulation approved and adopted
From private Jupyter notebooks to 3,000+ neuron simulations in one month.
Read the spotlight → - 03
A design system for scientist-builders
A living design system that scientists, product owners, engineers and AI can all build with.
Read the spotlight →
