Harry Wixley's Projects
Theorem proving in Isabelle
A python script to transform images into ASCII art in your terminal. Nick Cage DLC included.
Using DNA sequence alignment tools (ie. BLAST) to detect the existence of the GULO gene (used to biosynthesise vitamin-C) in varying different organisms.
Using DNA sequence alignment tools (ie. BLAST) to detect the relation between SFARI genes and Autism Spectrum Disorder (ASD).
Software engineering project: creating a customer & employee client for a bike rental company.
A CLI for talking to ChatGPT, saving conversation logs, and turning code blocks (language agnostic) from it's responses into executable shell commands.
Asymmetric encryption, email spoofing, mitm attacks
Using buffer overflows to exploit vulnerable C programs.
Preprocessing the STOIC2021 dataset for detecting COVID-19 severity
A drone air-quality mapping system. The drone's movement is constrained to moving in fixed increments, and only angles of 10. The system retrieves drone air-quality stations, and no-fly-zones as Geo-JSON objects from a webserver. The system then uses these to find an optimal route to pass through all the stations without going into any no-fly-zones
iOS App for monitoring and sending commands/message to an autonomous door handle sanitisation robot
A study of the problem of overfitting in deep neural networks, how it can be detected, and prevented using the EMNIST dataset. This was done by performing experiments with depth and width, dropout, L1 & L2 regularization, and Maxout networks.
Commercial iOS fall detection app. Connects to a Polar H10 device for triaxial acceleromter and ECG signals. These signals are passed to a trained ResNet152 model using Tensorflow background processes for live inference.
A localhost server to ensure secure and private data collection for fall detection data. This server uses IP-based whitelisting for security. This server streams ECG & Accelerometer data in chunks using a circular buffer to mitigate data loss.
An iOS fall detection data collection system. It uses CoreMotion for retrieving accelerometer, magnetometer, and gyroscope sensors, and interfaces with a PolarH10 chest strap for ECG data using the Polar SDK.
Preprocessing my fall detection dataset using data standardisation and sliding windows, and splitting this data into train/validation/test sets. Modelling performed on PyTorch using LSTM and CNN networks. The final models were exported to `.tflite` files to be run on a mobile phone. The best performing model was the ResNet152 with 92.8% AUC.
Turning rational numbers into fractional forms using concepts from the Farey algorithm.
Language identification and PP-attachment disambiguation
Supervised and semi-supervised training of Hidden Markov Models using the Viterbi algorithm
Model database
An automated workflow for generating visualizations of my GitHub stats
A very simple file server in Go
Introductory Applied Machine Learning - Coursework 1: simple machine learning model development, and analysis using Sci-Kit Learn and Numpy
Introductory Applied Machine Learning - Coursework 2: development and analysis of machine learning models on the FashIonMNIST dataset using SciKit-learn
Informatics Large Practical - Coursework 1: Rendering air-quality maps by parsing air-quality geo-json data from a local webserver.
Informatics Large Practical - Coursework 2 Report: Developing pathplanning software for a drone and mapping it's route on a map that avoids no-fly-zones parsed as Geo-JSON objects.
Task1 - data analysis & classification with multivariate Gaussian classifiers
Simulator of a MIPS processor in C: executes 32 bit binary instructions, including dynamic memory allocation.