The Challenge

BenevolentAI's mission to revolutionise drug development focused on early-stage discovery where costs of failure is relatively low but impact on clinical success is high.

The company built an extensive knowledge graph capturing current biological knowledge and deployed AI models to predict novel drug hypotheses.

The problem of evaluating AI outputs was already solved with the product called Target Triage.

What remained was the challenge of selecting AI inputs — knowledge graph entities to drive AI models. Historically, entity selection was ran by bioinformaticians who provided multi-day data exploration services to drug discoverers and configured AI runs. This process was manual and slow, suffering from human bias and error, and often failing to capture the underlying molecular mechanism of disease.

To realise the benefits of AI-powered drug discovery, entity selection had to become faster, reliable, and precise.

My Approach

First, I led design-driven discovery to understand the problem and scope the solution in collaborate with product directors. Semi-structured interviews with drug discoverers and bioinformaticians, plus observing the legacy process, enabled first wireframes within days.

Wireframing and prototyping came next. I took the team through rapid iteration rounds, consulting with tech teams and validating with users.

In parallel, I hired my design team. Once the product workflow was agreed, we split feature ownership between designers. I designed the Venn diagram, network visualisation, and the ontology tree.

The Solution

Entity Explorer is a web app that gives non-technical users ability to browse thousands of knowledge graph entities, compare entity benchmarks, explore ranked data-driven recommendations, visualise gene networks, and create entity shortlists for evaluation and AI model input.

The product forms the Benevolent Platform supporting end-to-end AI-powered early-stage drug discovery.

With other Benevolent Platform apps, Entity Explorer uses the Benevolent Design System, captures detailed user analytics (Mixpanel), and was deployed to BenevolentAI's commercial collaboration with AstraZeneca.

The Result

Entity Explorer gave drug discoverers self-service access to knowledge graph entities, letting them select AI inputs with no dependency on bioinformaticians (who could now focus on valuable R&D work).

This accelerated entity selection from days to hours and allowed BenevolentAI to build a pipeline of 26 programmes in 4 years (including 5 targets selected by AstraZeneca).

Entity Explorer made entity selection more reliable and robust. It eliminated the risk of human error, reduced bias, and helped teams understand disease biology to target underlying mechanisms with precision and confidence.

The network visualisation was used in the BenevolentAI-AstraZeneca collaboration. AZ scientists praised its ease of use, interactive gene pathway exploration, and beautiful interface.