// selected work

Projects that earned their keep.

A selection of API integrations, AI agents, models, and automations I’ve built. The problem, what I shipped, and what it changed.

  • API integrations
  • AI agents
  • Models
  • Automation
live demo · predicted profile NACA 23012
Reynolds no. 1.0×10⁶
Velocity 50 m/s
Angle of attack 4.0°
FEATURED Neural network · model

Aerofoil-shape prediction network

The problem. Choosing an aerofoil for a given set of flight parameters normally means hours of an engineer’s time.

What I built. A pretrained PyTorch neural network that, given specific flight conditions, predicts the ideal aerofoil shape. I built the dataset from scratch, designed and trained the model, fine-tuned it for accuracy, and used GPU acceleration with CUDA. It returns the top candidate profiles ranked and ready to plot.

7ms
run time
GPU trained
using CUDA
Custom
built dataset
PyTorch CUDA NumPy Pandas
json_to_csv.py 
resp = requests.get(API_URL, timeout=10)
resp.raise_for_status()  # errors on 4xx / 5xx
rows = resp.json()["results"]

with open("export.csv", "w", newline="") as f:
    writer = csv.DictWriter(f, fieldnames=FIELDS)
    writer.writeheader()
    writer.writerows(rows)
API integration

Database-to-CSV export pipeline

Built for a leading Canadian university: reads records from a database through its API and converts the JSON to correctly formatted CSV, with field validation and robust error handling so nothing slips through silently.

Python Requests CSV
casualnavy overshirtwhite teeslim chinos
smartcharcoal blazeroxford shirttailored trousers
sportblue hoodietech teejoggers
Computer vision · neural network

Wardrobe recommender

Point it at a photo of yourself and it recommends a full outfit to wear. A neural network reads the image — looking at colours, tones and garments — and picks matching items from your wardrobe. Built and run entirely from the command line.

Claude API CNN Pillow
Embedded systems · radio

Rocket telemetry & drop predictor

A Raspberry Pi built to fly in a rocket and radio its live location back to a ground laptop. I wrote the software for both ends of the link. A companion program models the flight and predicts where seeds released mid-air would drift and land.

Python Raspberry Pi Radio

// let’s build

Have something you’d rather software did?

Tell me the tedious thing. I’ll tell you how I’d automate it. Replies within a day.

Send an enquiry → About me