Projects
Finger splint (Fusion360)
Self-directed CAD learning project: designed an improved splint for boutonniere deformity (a common tendon injury causing PIP joint deformity) by identifying comfort and durability issues in existing designs (Oval-8, buddy loop, Capener, popsicle). Iterated through three design variations to balance structural support with wearability, arriving at a dual-arch geometry that maintains joint alignment while reducing bulk. Selected nylon over PP and ABS based on flexural modulus and low creep under sustained load, which is critical for a device that must hold constant corrective force through a full day of wear. Produced a manufacturable 3D model and engineering drawing, sized for S/M/L based on percentile finger anthropometrics.
Skills demonstrated: Fusion 360 (sketching, multi-face modelling, rendering, engineering drawings), materials selection, trade-off analysis and iterative design.
Designing a bioreactor (Group engineering challenge)
Designed a bioreactor system to produce a yeast-based subunit vaccine for tuberculosis, targeting a real-world deployment scenario in Uganda. As a team, we defined a target population (93% of Uganda, excluding those with yeast allergies or HIV), designed the heating, mixing, and cell-monitoring systems for a scaled-down 24-hour prototype run, evaluated potential stakeholders, and produced a feasibility assessment covering facility sizing, capital costs and cost per dose.
My contribution, The Impedence Subtask: I designed and built an impedance-based probe to monitor yeast cell growth, so the team would know the optimal moment to harvest cells for vaccine production. This involved experimentally optimising electrode placement, input frequency, and amplifier gain to maximise sensitivity to changes in cell concentration, then building a calibration curve relating probe output to yeast density. The final probe design (6 kHz input frequency, gain of 100 and probe position as shown in the diagram on the left) successfully distinguished yeast solution from buffer across a range of concentrations and was used to track cell growth during the team's 24-hour bioreactor run, correctly identifying the transition into stationary growth phase.
Skills demonstrated: experimental design, signal processing and calibration, working with lab electronics (function generator, oscilloscope, differential amplifier) and data analysis under real-world measurement noise and uncertainty.
CupCount sensor watch (medical device prototype)
Designed and tested a wearable device that monitors ambient temperature and heart rate to deliver personalised hydration reminders. Built and validated a working prototype through four targeted tests: sensor consistency over time, temperature accuracy across environments, internal LED interference, and optimal heart rate sensor placement on the body. Established activity-and temperature-based thresholds to trigger reminders, requiring both sensors to agree before issuing urgent alerts, reducing false positives from normal movement or ambient warmth. Evaluated the trade-offs between false positives and false negatives in order to prioritise user safety.
Skills demonstrated: calibration testing, embedded prototyping (Circuit Playground, 3D-printed enclosure), soldering and electronic assembly, threshold-based decision logic and design iteration from testing.
OralSight (Tounge cancer screening app)
Built an AI-trained app to screen for early signs of oral cancer. We identified a market need for more accessible oral health screening tools, particularly for tongue cancer, where earlier detection can significantly improve outcomes. We therefore designed a tool to help identify potential signs of oral cancer at an earlier stage, shifting the focus from treatment to early detection and intervention.
Using MATLAB, we trained an image-classification model (built on a fine-tuned pretrained AlexNet) to detect four visual conditions: ulcers, leukoplakia, hairy tongue, and strawberry tongue. The pipeline involved organising and labelling a tongue-image dataset, splitting it into training and validation sets, applying data augmentation (rotation, reflection, scaling) to improve generalisation, and evaluating performance with a confusion matrix.
The model was integrated into a full app concept: users complete a lifestyle risk questionnaire, receive guidance for capturing an accurate tongue photo and get an personal generated risk percentage of oral cancer. To manage the ethical risk of false positives, the app is explicitly framed as a screening and awareness tool rather than a diagnostic one, with built-in referral to professional care above a set risk threshold.
Skills demonstrated: MATLAB, dataset preparation and augmentation and translating a technical model into a health product.
Watch our advert!
FEA of a femur following a lateral fall (SolidWorks)
Tasked with a lateral fall simulation on a femur bone in SolidWorks. The femur geometry was imported, appropriate fixtures were applied (fixed support at the base, roller fixtures at the trochanter), custom material properties were set, and a mesh was generated. A fall-impact load of 9656 N was calculated from body height, mass, tissue stiffness, and gravity, then applied at the femoral head.
The resulting stress and displacement plots were evaluated and linked back to real-world context: peak stress (1.77×10⁷ Nm⁻²) occurred at the load point, with a secondary concentration at the femoral neck, while displacement peaked at the load point and dropped to zero at the fixed base, reflecting the bone's stiffness. A safety factor of 6.7 was calculated by comparing the material's failure stress to the applied stress, consistent with the physiological safety factor typical of bone (~6), indicating the femur would be unlikely to fracture under this loading. The femoral neck was identified as the most probable fracture site, matching clinical observations from further research.
The report also assesses the suitability of the model's assumptions (homogeneous material, point loading, static analysis, idealised fixtures) and proposes improvements for a more realistic simulation.
Skills demonstrated: importing files into SolidWorks, setting different types of fixtures, setting custom material properties, producing plots in SolidWorks, calculating loads in real-world contexts, evaluating both model suitability and simulation results and linking numerical results to literature.