Cities need street-level, real-time environmental intelligence to protect public health and plan for climate resilience. Leading the development of the EU-funded UrbanAIR Digital Twin as a Service (DTaaS) platform, delivering real-time, street-level air quality and heat resilience forecasts across three European cities, in collaboration with international partners including VITO (Belgium) and the Barcelona Supercomputing Center.

Large-scale ML workflows are often bottlenecked by fragmented, heterogeneous data sources. Leading research on data virtualization services for large-scale ML workflows and service composition & discovery—including LLM-based automated metadata and keywords generation in high-volume data virtualization services, and process-aware knowledge graphs for agent-driven service discovery and grounded execution.

UK infrastructure research increasingly depends on modelling and analysis at unprecedented scale. DAFNI is a national computing platform that enables researchers to run advanced simulations across transport, water, energy, and city-scale systems. Led the development of a cross-platform data bridge connecting DAFNI and CEDA (Centre for Environmental Data Analysis) services to accelerate environmental modelling and streamline researcher access to national data resources.

Led development of LLM-based semantic search, information retrieval, and automated metadata generation at STFC, including (a) grant proposal search and (b) semantic search, clustering, and topic modeling of scientific data from STFC's DataGateway.

Applied machine learning to hyperspectral imaging for crop disease prediction, in collaboration with RAL Space.

During COVID-19, epidemiologists needed to make high-stakes decisions from tens of thousands of rapidly evolving datasets, with no existing infrastructure to visualize or search them at that scale. Initially volunteered as a software architect for the Royal Society's RAMP initiative, where a working system was needed within three months. To meet this deadline, ran an agile process built on continuous, direct stakeholder engagement instead of fixed sprints. Later, funded by EPSRC, led development of RAMP-VIS, a visual analytics infrastructure, in collaboration with the UK visualization research community and the Scottish COVID Response Consortium (SCRC). Designed and implemented infrastructure for rapidly deploying visualizations, dashboards, and search capabilities across tens of thousands of heterogeneous, continuously evolving datasets, and developed novel methods for semi-automated visualization generation.

The SKA is the world's largest radio telescope, generating astronomical data at a volume and rate where real-time streaming and visualization are essential to monitor and act on data as it arrives. Led the design and development of a scalable, real-time streaming and visualization infrastructure to meet this need, with a low-latency data streaming architecture and web-based rendering pipelines for high-volume data streams.

Horus Security Consultancy's intelligence services depend on large-scale open-source data, but manual workflows could no longer keep pace with the growing volume of online sources. In this Innovate UK-funded university-industry collaboration, led development of a prototype system for scalable, automated open-source intelligence gathering: worked with analysts to capture operational workflows and translate domain requirements into system architecture, and designed and implemented an end-to-end platform for data ingestion, management, search, and provenance, providing analysts with efficient access to information. Later joined Horus as Lead Data Scientist to lead the transition of the prototype from proof-of-concept to commercial software, building and leading a team of software engineers to deliver multiple customer-facing products.

Seismologists at the International Seismological Centre (ISC) previously relied on inefficient, paper-based workflows to review complex seismic events across multiple data types. Developed VBAS—an interactive visualization system that integrates diverse seismological data, including hypocentres, magnitudes, phase arrivals, travel-time curves, seismicity maps, station geometry, and more. VBAS provides a unified interface that helps ISC analysts detect patterns, identify anomalies, and perform data analysis far more efficiently. The system replaced the ISC's 30-year-old paper-based process.

Laing O'Rourke, an international engineering and construction company, holds a vast archive of project knowledge whose reuse and monetization depend on searching it quickly and reliably - beyond what standard enterprise search could provide. Developed a proof-of-concept visualisation-assisted search engine to address this challenge: a focus+context visualization for rapid observation of false positives and false negatives in search results, a Search Provenance Graph with visual representations of search collaboration, and a special-purpose enterprise search engine with machine learning of ontologies to improve search precision and recall.

Oracle's database runs mission-critical operations for customers worldwide, where defects or missing features can disrupt business operations - making reliable engineering essential. Developed and debugged Oracle database service features, and independently identified the need for, designed, and built automation tools that improved efficiency and reduced cost across the patch build, test, and delivery pipeline.

Commercial real-time operating systems (RTOS) added significant licensing cost to industrial automation controllers, motivating ABB's investigation of Linux as a free, real-time-capable alternative for these time-critical embedded systems. Developed Linux system software for ABB's industrial automation controllers and conducted performance benchmarks across various Linux RTOS flavours to assess real-time suitability.