Projects
Test systems, instrumentation, and research programmes — from a customer-sponsored R&D engagement for aerospace and defence organisations, down through the fixturing that holds a test article steady enough to measure, to the solvers underneath the simulations.
Build
Mechanical design and fabrication, high-voltage drive and diagnostics, force balances, motion control, and NI data acquisition — specified, built, and commissioned.
Measure
Experimental planning, measurement-method selection and auditing, automated acquisition and processing in Python and MATLAB, empirical and reduced-order modelling.
Report
Translating objectives into research plans and deliverables, running customer-facing technical meetings, and writing the reports that let other people act on the result.
01 Engagement
The shape of it
A customer arrives with a programme objective, not a test plan. This engagement was largely about the distance between the two — turning objectives into research plans, experiments, work assignments and a schedule of deliverables, then owning the technical relationship week to week.
I led weekly technical meetings with customer engineering teams: presenting results, resolving open technical questions, and agreeing what happened next. I planned research, testing, modelling and analysis across a small team, delegated where it made sense, and integrated the individual pieces into project-level outcomes.
I owned the biweekly and monthly technical reports — the documents that synthesised experimental data, modelling results, literature, technical risks and recommendations into something a customer's engineering organisation could act on.
On the bench I contributed to high-voltage test-system development, plasma-actuator characterisation, instrumentation troubleshooting, data analysis and reduced-order modelling.
The result I am most pleased with
A controlled comparison of Rogowski-coil and resistive-shunt current measurement on the same discharge surfaced major negative-cycle artefacts — the class of systematic error that quietly biases an entire dataset when nobody thinks to check the instrument against itself. It became a recommendation to change how discharge current was characterised. Nobody asked for that study; it came out of not trusting a number that looked slightly wrong.
02 Test systems
A modular NACA 0018 airfoil carrying an array of surface actuators, sized for a 3′ × 3′ wind-tunnel test section, with a force balance measuring lift, drag and pitching moment across an adjustable angle of attack.
I advised the senior team that designed and validated it: defining technical requirements, running design reviews across mechanical design, high-voltage integration, instrumentation and tunnel implementation, then guiding the testing, troubleshooting, data interpretation and final validation.
A first-generation platform for evaluating surface actuators on an aircraft geometry: a scaled model split at the vertical nose-to-tail plane so a larger wingspan would fit the available test section, with modular wing and tail mounting, a rigid structural body, and supporting structure for either in-tunnel or stand-alone use.
I contributed mechanical design, high-voltage system integration, CFD in Ansys Fluent, experimental planning and project coordination. Testing exposed the mechanical and measurement limits that set the requirements for the in-tunnel system built after it — which is arguably the more useful outcome.
03 Automation
Manual data collection is where experimental campaigns go to die: it is slow, it is where operator variance creeps in, and it is why interesting parameter sweeps do not get run.
I automated the experimental procedure end to end — multi-axis CNC motion control driving the probe, National Instruments acquisition capturing the channels, and processing that took raw records to results without hand editing. Collection and processing time dropped by roughly 63%, which mostly bought back the ability to run sweeps that had previously been unaffordable.
The same approach carried into doctoral work: CNC positioning stages, live thrust-and-drag measurement, moment-isolation hardware, and purpose-built test articles and mounts, all designed for repeatability and throughput rather than a single run.
NI DAQ, oscilloscope-based voltage and current diagnostics, intensified high-speed imaging, precision force measurement.
Multi-axis CNC positioning for velocity-profile traverses and repeatable actuator placement.
Python and MATLAB pipelines: signal processing, batch analysis of large experimental datasets, empirical model fitting.
Moment-isolation hardware, force balances, airfoil test articles and mounts designed in SolidWorks and machined in-house.
04 Computational
Graduate coursework, and I label it as such — but the distinction that matters is between running a commercial code and implementing the numerics underneath it. These are the latter.
Electrohydrodynamic flow is a multiphysics problem: an electric field ionises the air, the ions drift, and their momentum transfer to neutral gas drives a jet. Solving it means coupling Poisson's equation for the field to Navier–Stokes for the flow through a Coulombic body-force term.
I built that solver — separate Laplace and Poisson field solvers, a Navier–Stokes component, and the coupling between them — and then checked it against the published analytical point-to-ring corona model of Guan et al. (2018), because a solver you have not compared against a known answer is just an opinion with a plot attached.
The point-to-ring geometry is not incidental. It is the electrode configuration behind the axisymmetric plasma synthetic jets I now characterise experimentally, so this was the same physics approached from the other side.
A projection-method incompressible Navier–Stokes solver written in Fortran 90, with Python implementations alongside it, plus a standalone Poisson solver and 2D orthogonal grid generation written in MATLAB.
The part I would point at is the rotating cylinder. I solved it with my own projection-method code across five successive mesh refinements, and solved the same case in Ansys — so the comparison was between an implementation I understood completely and a commercial code I did not. That is a small version of the exercise I think every simulation claim should be able to survive.
Projection method for the incompressible equations, structured grids, successive mesh refinement.
2D orthogonal grid generation implemented directly rather than taken from a mesher.
Fortran 90 for the solver core, MATLAB and Python for meshing, prototyping and post-processing.
Own implementation run against Ansys on the same case, and against analytical solutions where they exist.
05 Breadth
The fair question to ask someone who has spent five years on one problem is whether they can work on anything else. These are graduate courses I took deliberately outside my field, each ending in a written report and a talk.
2025ME 567
Molecular engineering · report and seminar presentation
2023ME 537
Mechanical engineering · report and presentation
Neither is my research. That is the point of listing them — picking up an unfamiliar literature quickly and reporting on it honestly is most of what a technical engagement actually consists of.
06 Ongoing
The longest-running project on this page, and the one that taught me most of the rest of it: a multi-year experimental campaign characterising how actuator geometry, dielectric properties, voltage and frequency determine what a surface discharge does — assembled into operating maps rather than isolated data points.
Three publications and international conference presentations have come out of it so far, alongside mentoring undergraduate researchers and advising capstone teams. I also serve as the lab's Chemical Hygiene Officer, where a sustained push on procedure and documentation moved the EHS inspection score from 44 to 96.
07 Contact
If you have a measurement that is not behaving, a test system that needs building, or results that need turning into something your organisation can act on — I would like to hear about it.