👉 Discover how Cambridge Spark helps organisations build the data and AI capabilities needed to turn strategy into measurable impact: cambridgespark.com

This week, Dr Judit Guimera Busquets, Head of Data Science at Datasparq, joins Dr Jeremy Bradley to trace the journey from her PhD on air traffic network forecasting through to leading data science teams delivering real-world AI projects.

Judit explains why forecasting inside a complex network is fundamentally different from standard demand prediction: when a single airport pair is removed, the cascade effect ripples across an entire system. She walks through the multi-stage modelling framework she developed, covering city pair demand generation, network evolution, itinerary assignment, and long-term scenario planning.

The conversation then turns to what actually happens when structural shocks like a pandemic break a model's core assumptions and why human-in-the-loop design is not optional. Judit also sets out what she looks for in data scientists: pragmatism over perfection, simplicity over complexity, and a production-first mindset from day one.

She closes with her view on where applied AI is heading, including the rise of small, fine-tuned specialist models and why AI governance remains the most overlooked challenge in the field.

Follow Data & AI Mastery on Apple Podcasts, Spotify, or YouTube to stay ahead of the algorithm.

If you enjoyed this episode, why not check out the Data & AI Mastery episode with Richard Masters, VP of Data and AI at Virgin Atlantic. You will learn more about how the airline leverages AI and data-driven strategies to enhance operations, optimise pricing, and deliver premium customer experiences:

Apple: https://podcasts.apple.com/gb/podcast/mastering-data-ai-insights-from-virgin-atlantics-vp/id1779783413?i=1000697801419

Spotify: https://open.spotify.com/episode/1MY3AZCvDr5eS1HlfBudHy?si=ca5ec9b2fe6d44b1

YouTube: https://www.youtube.com/watch?v=DQ3mTwTzJvA

Glossary Terms

Hub-and-spoke Model: a centralised organisational architecture where a central core connects to multiple peripheral nodes. Traffic, communication, or inventory flows through the hub rather than directly between spokes.

Network Theory: a multidisciplinary framework used to analyse complex systems by representing them as mathematical graphs

Econometrics: the application of statistical and mathematical models to economic data

Human-in-the-Loop: a collaborative AI approach where humans actively participate in an automated system's training, refinement, or operation.

Linear Regression Model: a fundamental statistical and machine learning algorithm that models the relationship between a dependent variable and one or more independent variables by fitting a straight line to the data.

Chapter Markers

(00:00) - What makes network forecasting different from standard demand prediction

(05:54) - How historical data fails when the network itself evolves

(10:05) - Modelling link addition and removal as classification problems

(13:26) - Designing for medium and long-term policy evaluation, not daily operations

(17:57) - What happens to a model when a structural shock like a pandemic hits

(22:22) - Human-in-the-loop: adjusting elasticities and running what-if scenarios

(27:20) - What great data scientists actually look like in a consulting environment

(30:03) - Getting stakeholders to use AI: champions, end users and change readiness

(32:00) - Where applied AI is heading: small specialist models and the governance gap

Useful Links

Connect with Dr Judit Guimera Busquets on LinkedIn: https://uk.linkedin.com/in/judit-guimera-busquets-696ab74a

Learn more about Judit’s PHD here: https://openaccess.city.ac.uk/id/eprint/24689/1/Busquets%2C%20Guimera.pdf

For more AI insights follow Jeremy on LinkedIn: https://uk.linkedin.com/in/jeremy-bradley

Explore Cambridge Spark’s AI upskilling programmes at https://www.cambridgespark.com

👉 Discover how Cambridge Spark helps organisations build the data and AI capabilities needed to turn strategy into measurable impact: cambridgespark.com

Planning healthcare at scale is one of the most consequential challenges facing the NHS. In this inaugural episode of our new show, Inside The Algorithm, Cambridge Spark’s Chief AI Officer, Dr Jeremy Bradley, is joined by Dr Lucy Morgan, Head of Simulation at the NHS Strategy Unit, to explore how simulation modelling and AI are being used to tackle that challenge head-on.

Lucy led the development of a hybrid simulation model for renal replacement therapy, which began with the Midlands Kidney Network and is now being rolled out nationally. She breaks down how the model works, combining system dynamics with discrete event simulation to project future demand and model patient pathways through dialysis and transplant.

The conversation moves into the harder questions: how do you validate a model in a live healthcare environment, how do you rebuild trust with new organisations who weren't part of the original build, and where does AI fit into making models more portable and reproducible?

Lucy also shares her current research interest in using AI agents to help NHS staff query and interrogate analytical models directly, without needing deep modelling expertise themselves.

A rich, technically substantive episode for anyone working at the intersection of AI, simulation and real-world deployment.

Follow Data & AI Mastery so you never miss a future episode of Inside The Algorithm.

Liked this episode? Why not listen to the conversation Dr Raoul Gabriel Urma had with Ming Tang from the NHS on the Data & AI Mastery podcast:

Apple: https://podcasts.apple.com/gb/podcast/transforming-healthcare-through-data-ai-human-centred/id1779783413?i=1000736381325

Spotify: https://open.spotify.com/episode/27wqHT0dnRMwKMedj2uoTw?si=5aabd4015dfa425d

YouTube: https://www.youtube.com/watch?v=JERGmZUumvA

Glossary Terms

Model reproducibility: the ability to obtain consistent computational results using the same input data, code, and environment

Neural Network: a method in artificial intelligence that teaches computers to process data in a way that is inspired by the human brain

Sequential Hybrid Simulation Model: a modeling approach that combines two or more distinct simulation paradigms (such as System Dynamics, Discrete-Event Simulation, or Agent-Based Simulation) by running them in a specific order, where the output of the first model acts as the input for the next

System Dynamics: a computer-aided modeling methodology used to understand, analyse, and manage complex, dynamic systems over time

Stock-flow System: a foundational system dynamics modelling concept mapping how quantities (stocks) accumulate over time, influenced by rates of change (flows)

Chapter Markers

(00:00) - Opening: The case for AI-assisted simulation in the NHS

(02:11) - About the NHS Strategy Unit and the backwards consultancy model

(05:04) - Tackling NHS waiting lists with queuing theory

(10:26) - Complexity of scheduling: theatre capacity, rotas and prioritisation

(13:42) - Introducing the renal replacement therapy planning model

(18:32) - Data sources, model calibration and validation

(23:00) - Rolling the model out nationally: trust, workshops and participative modelling

(26:47) - A call to the community: sharing best practices in model reproducibility

Useful Links

Connect with Dr Lucy Morgan on LinkedIn: https://www.linkedin.com/in/lucy-morgan-97072041/

For more AI insights follow Jeremy on LinkedIn: https://uk.linkedin.com/in/jeremy-bradley

Explore Cambridge Spark’s AI upskilling programmes at https://www.cambridgespark.com