News

Over the past few months, the LEAP team has welcomed intern Lilian Sudi through the HDR UK Black Internship Programme. Working closely with Researcher in Residence Miquel Perello Nieto, Lilian has been developing a partner matching tool designed to foster new connections across the LEAP network.

We caught up with Lilian to learn more about the project and hear about her experience as part of the team.

I applied to the HDR UK Black Internship Programme (part of the 10,000 Black Interns Programme) seeking practical experience in machine learning, something I had only explored within the confines of my classroom.

The BIP promised a structured pathway with support, training, and genuine mentorship. Nearly 80% of interns in the programme say they wouldn’t have landed an internship without it. For me, it opened a door I didn’t even know existed.

That door led me to LEAP Digital Health Hub.

LEAP brings together five universities across South West England and Wales, alongside researchers, clinicians, entrepreneurs, NHS leaders and social care organisations spanning the region – from major cities to remote rural communities. Each team is brilliant in its own domain, yet often isolated in its impact.

Because most of LEAP operates remotely, spontaneous connections are rare. You can’t just bump into a researcher from Exeter if you’re based in Wales. Without intentional spaces for collaboration, these conversations simply don’t happen.
And so, the same questions keep surfacing:

• Who else is working on this?
• Who needs what I have?
• Where is my perfect collaborator?

That’s the gap we’re trying to close.

I joined LEAP in June 2025. The interview was fantastic, I got to meet Hanna Isotalus and Miquel Perelló Nieto, and I was immediately drawn to the team’s energy, creativity, and genuine passion for what they were building. On my first day, that same energy filled the room.

Image 1 (Miquel & I): With Miquel Perelló Nieto, ML researcher at University of Bristol and mentor for the LEAP partner matching project.

From the get-go, I was introduced to the challenge: help the right people find each other. And no, they weren’t talking about Tinder.

Typically, when you join an organisation, you make friends, ask around, and slowly figure out who’s working on what and where collaboration might happen. It’s tedious, time-consuming, and inevitably, you miss out on connecting with half the brilliant people in the room.

Our solution? Build a system that makes those connections seamless, one where you can search, find, and connect with people whose skills match your needs instantly. This is essential for a research institution.

Sure, generic platforms exist: LinkedIn, collaboration hubs, and professional networks. But they often prioritise volume over meaning. They match keywords, not complementary expertise. They’re built for scale, not for the messy, human work of digital health, where a behavioural scientist, a UX designer, and an NHS manager might form the perfect team, but algorithms rarely see that.

The other option is to rely on serendipity, conferences, casual conversations, and networking events. But for every connection that happens by chance, dozens of ideal collaborations never do.

Through building this platform, I’ve come across people I’d never met before, individuals I might never have known existed within LEAP. And as these profiles emerged, I realised something important: brilliant people cluster in patterns. Not by job title. Not by keywords. But by the problems they solve, the skills they bring, and the gaps they fill.

Under the guidance of Miquel Perelló Nieto, a machine learning researcher at the University of Bristol, we’ve been developing a system that uncovers those patterns.

The Technical Approach:

We started with a traditional keyword-matching approach using TF-IDF (Term Frequency-Inverse Document Frequency).

It’s fast, straightforward and finds exact word matches between profiles. But here’s the problem: if one researcher describes their work as “sustainable infrastructure” and another as “renewable energy systems,” TF-IDF doesn’t recognise they’re actually working on the same thing. They’re discussing similar concepts, but using different language. The algorithm misses the actual meaning.

That’s where semantic search comes in. Instead of just matching words, we needed to understand meaning. We therefore proceeded to work on a Semantic Search option as well.

Image 2 (LEAP Profiles UI): LEAP Profiles interface with dual matching modes – Explainable (TF-IDF) for transparent keyword matching and Semantic (SBERT + Cluster) for deeper expertise alignment across

1. Semantic Search (SBERT + Clustering): This method is fairly advanced. We employ Sentence-BERT embeddings to understand the deeper meaning behind each person’s expertise and research interests, not just surface-level keywords. For instance, “renewable energy systems” and “sustainable infrastructure” are semantically similar even if they don’t share exact words. We then use K-means clustering to segment LEAP members into research communities, enabling faster and more accurate matches within thematically aligned groups.

Image 3 (Search Results): Semantic search results displaying matched researchers ranked by similarity score with star ratings and cluster assignments – Professor Raul Santos-Rodriguez shows 41.8% match (Fair Match) based on shared research interests in machine learning and healthcare.

2. Explainability (SHAP): We know that researchers need to understand why two people are matched, not just that they are. Using SHAP (SHapley Additive exPlanations), we highlight which specific skills, research interests, and challenges drive each match. When the algorithm recommends a collaborator, it shows exactly which words and concepts contributed to that recommendation. This builds trust in the system and helps people quickly assess fit.

Our goal is to enable LEAP members to create profiles that showcase their expertise, challenges, and vision. The algorithm identifies deeper connections, complementary skills, shared challenges, and strategic alignment within digital health. Because when the right people connect within LEAP, every collaboration becomes an opportunity for impact, and every partnership helps accelerate change across the region.

Image 4 (SHAP Explanation): SHAP text explanation showing how semantic matching works – highlighting key terms like ‘machine learning’, ‘healthcare’, and ‘systems’ that contributed to the 41.8% similarity score between.

We’re currently developing the pilot phase of the platform. If you’re interested in expanding your network and exploring new collaborative opportunities, we’d love for you to share your details with the Hub. Your profile might be exactly what someone out there is looking for.

Working with Hanna, Miquel, Rachel, and the entire LEAP team has shown me just how transformative strategic connection can be. I have been able to truly understand people’s needs, identify gaps, and design systems that unlock human potential.

Through the BIP programme, I had the opportunity to contribute to an organisation that values impact and innovation, and to put my machine learning skills to real use. It has been such a rewarding experience and a genuine confidence boost. I’m deeply grateful for this opportunity.

To anyone considering the 10,000 Black Interns Programme: if my time at LEAP is anything to go by, don’t hesitate, apply. You’ll gain not only experience but also a deep appreciation for working with a team that prioritises mentorship, growth, and authentic partnership.

Have any questions? Reach out to the Hub team.

Apply to host a LEAP-funded intern

The Hub is funding a small number of places on the HDR UK Black Internship Programme in 2026. Find out more and pledge to host an intern.

Apply for an internship

Applications are now open for the 2026 Black Internship Programme. Find out more and apply on the HDR UK website.

Visit the 10,000 Interns Foundation website to find out about other programmes.