Academic Projects

Academic Projects

Tools built around the person at work. Each one starts from a human factors problem.

Project Lead · INEN 6301 UX with AI · 2026

Check-Mate

A mobile app that suggests a short break when your fatigue checks say you need one, then measures whether the break helped.

Adaptive micro-breaksReaction-time testMobile web appUser testing
Video demo coming soonA two-minute tour of the app from sign-up to break history.
Read the case study

The problem

Fixed break schedules ignore how tired a person actually is. People either break too late, after errors creep in, or break when they don't need to and lose time.

How it works

  • Set up. Create an account and pick favourite break activities
  • Quick check. A sleepiness rating plus a reaction-time test
  • Break. One short activity of up to 10 minutes, matched to fatigue risk
  • Recover. Rate how you feel and retest, so recovery is measured

My role

  • Led the team through the literature review and the study methodology
  • Designed and built the user interface with teammate Nirman
  • Leading the final stages of development and preparing the user experiments

Why it matters

It turns microbreak research into something a worker can use on their phone. Every check and break is logged and can be exported as CSV, so the data feeds straight into analysis.

AI and Software, Dr. James Curry · Oct to Nov 2026

TaktArm Twin

A digital twin of the lab's UR10e robot arm. You tell it the job in plain English, see the motion previewed in 3D, and approve it before the robot moves.

UR10e digital twinCycle timeLocal LLMHuman in the loop
Video demo coming soonOne command, from typing to approval to robot motion.
Read the case study

The problem

Programming a cobot for each new task is slow, and operators often cannot see what the robot will do before it moves.

What I built

  • Plain-language commands read by an offline language model
  • A preview arm that shows the plan before anything moves
  • The fastest process within the 1000 mm/s safety limit, with the time it saves
  • Charts comparing cycle time, energy and peak power across speeds
  • Payload checks against the 12.5 kg limit and bin fit checks

Results in simulation

37 to 40%
shorter predicted cycle time
65.8 s → 39.5 s
six-object batch
≈ ⅓
less predicted energy per cycle

These are twin predictions. They will be checked against measurements on the real robot.

Human factors angle

The operator stays in control, because no motion runs without approval. A planned user study will measure workload (NASA-TLX), trust and usability (SUS).

Built with

Python, FastAPI, Qwen 2.5 3B through Ollama, Three.js with the official UR10e model, NumPy and OpenCV. About 9,700 lines of code and 83 automated tests. Hardware is a UR10e with a Robotiq Hand-E gripper and wrist camera.

Workplace safety · 2026

5S Analyzer for Workplace Safety

An AI system that checks live CCTV and uploaded photos for 5S and safety problems across six industries, then scores and reports them.

Computer vision5S auditsSeverity scoringReact
  • Configurable severity scoring for each violation
  • Daily, monthly and yearly reports with trends, exported to Excel and Word
  • Role-based access, live camera monitoring and violation tracking
Video demo coming soonA walkthrough of a live 5S audit.
Research tool · Guide

Garmin Forerunner 55 Participant Guide

A screen-by-screen guide that helps study participants set up their watch and export their wellness data as CSV files.

Research tool · Guide

Garmin Timestamp Conversion Guide

A click-by-click Excel tutorial that turns Garmin timestamp_16 values into readable UTC times, with an example workbook.

Product design · 2024

Rotary Tabletop for Wrist and Forearm Ergonomics

An adjustable tabletop with a rotary positioner to reduce strain during keyboard use. Top 10 finalist at the 2024 Cardinal Ideas Pitch.

Optimisation · Python

Vehicle Routing with a Driver Fatigue Model

A multi-objective routing tool built with OR-Tools and NSGA-II that balances delivery cost against driver fatigue.

Scheduling · Python

Permutation Flow Shop Scheduling

A PuLP and CBC model that studies the trade-off between tardiness and schedule nervousness.

Experiment · 2025

Keyboard Slope and Wrist Support Study

A 2×2 factorial experiment on keyboard slope and wrist support height, analysed with ANOVA and Tukey post-hoc tests. Published at AHFE 2025.

Interested in working together?

Please send me a note. I will get back to you at the earliest.