Projects
Overview
This page presents selected project domains and representative work that illustrate my experience in applied data science, machine learning, optimisation, and decision intelligence, across both academic and industry contexts.
Rather than cataloguing individual tools or isolated experiments, the emphasis is on how analytical methods are formulated, integrated, and operationalised to support real-world decision-making under uncertainty and constraints.
Across projects, my role typically spans:
Problem formulation from loosely defined, real-world decision contexts
Methodological and system design, combining ML, optimisation, and domain knowledge
Implementation, validation, and scalability considerations
Communication of results to both technical and non-technical stakeholders
Decision Support Systems (DSS)
AI-Driven Decision Support for Complex, Constrained Environments
A core theme of my project work is the design and implementation of AI-driven decision support systems (DSS) for environments characterised by uncertainty, multiple objectives, and operational constraints.
These projects move beyond standalone predictive models, focusing instead on decision-centric systems where analytics directly inform planning, prioritisation, and strategic choices.
Typical characteristics of these DSS projects include:
Formalising ambiguous, real-world problems into structured decision models
Identification and modelling of:
Decision variables
Hard and soft constraints
Multiple, often conflicting objectives
Sources of uncertainty and risk
Integration of:
Predictive machine learning models
Optimisation algorithms
Rule-based and domain-driven logic
Design of interpretable outputs to support trust and adoption by decision-makers
These systems have been applied to problems such as:
Operational planning and scheduling
Resource allocation under constraints
Scenario evaluation and trade-off analysis
Strategic decision support in large-scale industrial and energy systems
The overarching goal is to ensure that analytical outputs are actionable, transparent, and decision-relevant, rather than technically sophisticated but operationally disconnected.
Optimisation-Focused Projects
Large-Scale and Multi-Objective Optimisation Applications
Many of my projects centre on optimisation-driven decision-making, particularly in settings where classical exact methods alone are insufficient due to scale, uncertainty, or problem complexity.
Representative optimisation-focused work includes:
Large-scale linear, nonlinear, and mixed-integer optimisation
Multi-objective optimisation and Pareto-based trade-off analysis
Constraint handling techniques for infeasible or noisy real-world data
Hybrid approaches combining exact optimisation with metaheuristics
In several projects, optimisation models are embedded within broader DSS frameworks, allowing decision-makers to:
Explore alternative solutions under varying assumptions
Understand trade-offs between competing objectives
Evaluate feasibility, risk, and performance simultaneously
This work reflects a strong emphasis on practical solvability and interpretability, rather than purely theoretical optimality.
Machine Learning & Forecasting Projects
Predictive Analytics for Planning and Operations
I have designed and delivered machine learning and forecasting solutions intended to directly support planning, operational decision-making, and performance monitoring.
These projects typically involve:
Development of predictive analytics pipelines aligned with operational and business KPIs
Time series forecasting for short-, medium-, and long-term planning horizons
Use of machine learning and deep learning models to capture non-linear dynamics
Incorporation of uncertainty estimation and scenario analysis
A key design principle across these projects is that forecasts are decision inputs, not final outputs. As such, models are evaluated not only on predictive accuracy, but also on:
Stability under changing conditions
Interpretability and explainability
Usefulness for downstream optimisation or policy decisions
Results are communicated through clear reports, dashboards, and scenario summaries, enabling stakeholders to translate predictions into concrete actions.
Integrated ML–Optimisation Projects
Learning-Enhanced Decision Systems
Several projects explicitly integrate machine learning with optimisation, reflecting my broader research agenda of combining learning and decision-making.
In these projects:
Machine learning models are used to:
Forecast demand, cost, or system behaviour
Learn representations or surrogate models
Capture uncertainty and variability
Optimisation models consume these outputs to:
Recommend decisions
Allocate resources
Evaluate trade-offs under constraints
This integration enables:
Adaptive decision-making
Scenario-based planning
More robust and flexible decision policies
Such hybrid ML–optimisation systems are particularly valuable in large-scale industrial, energy, and infrastructure contexts, where decisions must be revisited as new data becomes available.
Data Intelligence (Text Mining & Bibliometrics)
Data-Driven Insight from Unstructured Information
In addition to numerical and operational data, I have led and contributed to data intelligence projects focused on extracting insight from unstructured textual data.
This work includes:
Text mining and topic modelling to identify emerging themes
Bibliometric and citation analysis
Collaboration and network analysis
Research landscape mapping and evidence synthesis
These projects support:
Strategic research planning
Policy-relevant analysis
Systematic literature reviews
High-level decision support based on large document collections
Emphasis is placed on reproducibility, transparency, and interpretability, ensuring that insights can be trusted and reused.
High-Performance & Scalable Computing
Computationally Intensive Analytics and Optimisation
Several projects involve computationally demanding models, requiring careful attention to scalability and performance.
This includes:
Large-scale optimisation problems
High-dimensional machine learning models
Parallel and GPU-accelerated computation
Efficient algorithmic design for real-world deployment
The focus is on achieving a balance between computational efficiency and decision quality, particularly in time-sensitive or large-scale applications.
Cross-Cutting Methodological Philosophy
Across all projects, a consistent methodological philosophy is applied:
Decision-first design, rather than model-first development
Tight integration of machine learning, optimisation, and domain knowledge
Emphasis on robustness, scalability, and interpretability
Clear articulation of assumptions, limitations, and trade-offs
This approach ensures that analytical solutions deliver practical value, not just technical novelty.