We build AI systems that forecast whether emerging healthcare technologies will succeed in the real world, not only whether they work scientifically, but whether they will be adopted, deliver impact, and capture value sustainably.
Our interdisciplinary team integrates AI/data science, technology management, and medical sciences, co-designing forecasting tools with clinicians and biological specialists to reflect real-world constraints and decision needs.
We are developing a cloud-based AI platform that analyses patents, publications, clinical trials, and social/health signals to forecast technical feasibility, societal impact, implementation barriers, and commercial viability of emerging healthcare technologies.
Healthcare innovation often fails in translation because decision-makers lack integrated, reliable signals about
Feasibility vs. adoption/value: a technology may be scientifically promising but commercially unsustainable or poorly adopted.
Implementation friction: clinical trial information is frequently difficult to navigate, slowing matching and adoption.
Equity gaps: high-impact solutions may be underfunded when market incentives are weak, worsening health inequities.
Our work targets these bottlenecks by making translation forecasting multidimensional, evidence-based, and scalable.
We model the innovation journey end-to-end through an iterative, four-stage AI framework.
We use Graph Neural Networks (GNNs) and transformer-based models to analyse large-scale bibliographic and patent data (e.g., PubMed, EPO) and estimate the probability of technical success.
We quantify societal impact using multimodal sentiment analysis and NLP, integrating demand signals and health indicators to forecast effects on disease burden.
We apply Named Entity Recognition (NER) to structure complex clinical-trial information and use reinforcement learning (RL) approaches to prioritise evidence in ways that reduce adoption friction for practitioners.
We evaluate commercial landscapes using approaches such as Word2Vec/Doc2Vec, LSTMs, and patent-citation trajectories to map technology mobility, differentiation, and competitive uniqueness.
A key element of our platform is a post-hoc correction and validation layer using multi-agent AI “debates”. Agents representing different viewpoints (clinical, technical, commercial) challenge and refine model outputs to improve robustness and reduce brittle recommendations. This structured critique helps convert raw analytics into stakeholder-aligned strategic recommendations, particularly for high-stakes decisions in pharmaceutical and public health settings.
We develop and test our methods in healthcare domains where translation is complex and data is fragmented, including oncology (with work involving radiopharmaceuticals and nanomedicine).
We aim to ensure our work supports societal benefit and reduces inequities by
Interdisciplinary co-design: predictions are triggered and evaluated with clinical and management expertise, not used in isolation.
Equity focus: we explicitly surface “high-impact / low-profit” technologies to support evidence-based recommendations for public intervention when market forces underprovide.
Lived-experience input: we incorporate feedback from public contributors in resource-limited settings (including Argentina) to ensure recommendations do not exacerbate existing inequalities.
Our six-member group brings complementary expertise across
AI & Data Science — Prof. Paul Yoo (BIDA+)
Technology Management — Dr Lourdes Sosa (LSE); Prof. Marcela Miozzo (KCL)
Medical Sciences / Oncology — Dr Shibani Nicum (UCL Hospitals)
Clinical & Biological Specialists / Public Contributors — Dr Yoana Vanni (oncology), Prof. Mario Rossi (molecular biology)
We collaborate closely across disciplines to integrate clinical expertise with AI modelling and strategy frameworks.
We work with academic, clinical, and international partners, including
Birkbeck Institute for Data Science and AI (BIDA+) & LSE Management — integrating ML models with technology strategy frameworks.
UCL Hospitals NHS Foundation Trust & UCL Cancer Institute — characterising scientific and adoption aspects in oncology.
International public contributors (Argentina) — incorporating feedback on technology use under resource-limited conditions.
Current support includes
NATO grant — supporting interdisciplinary collaboration and development of an explainable AI agent-based framework.
Public contributor grants — enabling specialist feedback from contributors in Argentina.