Jamie Benwell

The assembly line tracked hundreds of jigs and tools with paper checklists and spreadsheets — no real-time visibility, no audit trail, and frequent mismatches between what was recorded and what was actually on the line. I proposed a digital replacement on day one of my placement, then taught myself Oracle APEX and SQL from scratch to build it.

• Designed and built the full application end to end: QR-code scanning, live status dashboards, and automated fault alerts sent the moment a jig fails a check.
• Designed a custom authentication scheme to work around handheld-device limitations — later adopted as a recommended standard for other production apps across the business.
• Rolled the system out to the entire assembly line, working directly with 600+ associates and managers across every shift.
• Documented and trained a permanent team member to take over the app, so it outlives the placement.

A 5-inch class quadcopter build with a research core: designing custom propeller geometry from first principles rather than using off-the-shelf blades, then validating each iteration by print, test, and redesign.

• Two-window print strategy on a Bambu Lab P2S: a baseline carbon-fibre-reinforced print now, with rapid, low-cost PLA/PETG iterations during term time.
• Iterating on blade pitch and profile based on flight testing, not simulation alone.

Built without writing code in the traditional sense, using AI tools as the primary development method throughout. The app targets a specific gap I'd hit myself after years of casual app-based learning: vocabulary without real conversational fluency.

• Designed, built, and shipped independently — scoping, product decisions, and iteration all AI-assisted.
• Focused on spoken fluency and real usage rather than gamified vocabulary drilling.
• Learned to specify features precisely before generating them — every screen started as a written description of behaviour, not a vague prompt, because vague prompts produced unusable output.
• Built a working sense of where AI genuinely accelerates development (UI, boilerplate, iteration speed) versus where it still needs a human call (pedagogy, conversation pacing, what "correct" Swedish actually sounds like).
• That workflow carried straight back into my placement — I used the same AI-assisted approach to speed up dashboard and reporting work at Rolls-Royce.

Selected work from first and second year — included to show where the placement-year skills started.

Sim-Racing Wheel & Phone Mount First

Individual coursework, Year 2 (EGA266) — product design, from brief to manufactured part

Brief: design a computer peripheral that goes beyond aesthetics. I designed a phone-mount attachment for a Logitech G920 wheel that turns any racing sim into a live telemetry dashboard — benchmarked against real market products, then taken through concept generation, a weighted Pugh matrix, and iterative 3D-printed prototyping.

• Scored three concepts against a weighted Pugh matrix across 8 criteria, then combined the two winning concepts into the final hybrid design.
• Test prints failed twice — tolerance issues and fractured mounting points — redesigned wall thickness and fillets in response, not just resubmitted the same part.

2.2Nm torque withstood (spec: 2–2.5Nm) 269g total mass £7.49 build cost

FEA-Optimised Steel Bracket First

Individual coursework, Year 2 (EG264, Computer Aided Engineering)

Took a deliberately over-engineered bracket (safety factor 6.03 against a 4000N load) and optimised it down using a 16-run Taguchi design of experiments across four variables, validating every candidate with FEA before picking a winner — then stress-tested that winner at 60°C and reported honestly that it didn't hold up.

• Ran a mesh-convergence study first (16mm down to 4mm elements) so the FEA results could actually be trusted before comparing designs.
• Selected the optimal run by safety factor, stress, displacement and mass together, not by any single metric in isolation.
• Flagged a genuine limitation in the conclusion — the optimised design fails at 60°C — and proposed a specific redesign rather than hiding the result.

130g lighter than baseline SF 2.76 (target 2–3) 16 design iterations simulated

IMechE Design Challenge — Autonomous Repeatable Vehicle First

First-year team project, Group 22 (6 members) — Institution of Mechanical Engineers challenge

Build a vehicle that drives 4 metres to a wall and back within two minutes, as accurately as possible, using analogue electronics only — no microcontrollers, no digital logic. I modelled the tank-track drivetrain and running gear in CAD.

• Modelled the drivetrain and running gear — my CAD submission scored 100%.
• Built the stopping/timing logic around SR-latch relay circuitry, simplified from an initial transistor-based design after ambiguity over what components were permitted.
• Iterated wheel and shaft geometry through several laser-cut revisions to fix track slip and alignment.

Team of 6 Analogue electronics only 406.3g total mass

I build the things I think should already exist. At Rolls‑Royce, that was a tracking system now used by 300+ people every shift, saving the company £56k. At home, it's a propeller designed from first principles and an AI-built app that taught me more Swedish than five years of casual apps ever did. Final-year Mechanical Engineering, Swansea University — precision engineering on paper, motorsport in the blood.