Skip to content

Latest commit

 

History

History
40 lines (32 loc) · 3.63 KB

Imperial.md

File metadata and controls

40 lines (32 loc) · 3.63 KB
Raw

Imperial College London - Software Engineering Computing (MSc)

Welcome to my project list where I document all projects I've done as part of my degree at Imperial.

Machine Learning

1. Natural Language Processing 1 - Multilingual Psychothearapy Chatbot

Distinguished Project Award | Grade: 89% | TECH: Python (Pytorch, Numpy, Scikit-learn, Pandas, Transformers, NLTK), Slurm, Flask, React, AWS, Git

  1. Developed a multilingual emotion classifier with >90% accuracy in Mandarin and English using state-of-the-art language models (XLM-R, InfoXLM),
  2. Achieved a 6-fold reduction in inference latency and reduced cost of deployment through Knowledge Distillation (model compression),
  3. Developed a fluent, accurate and empathetic text rewriting system using generative language model GPT-2, trained using reinforcement learning and supervised learning methods, safe for use in psychotherapeutic treatment,
  4. Deployed ML models in a chatbot web application, with ~90% satisfaction ratings for engagement and usefulness in non-clinical human trials.

2. Natural Language Processing 2 - Patronising Language Detection

Coursework Grade: 90% | TECH: Python, Jupyter, Numpy, Pandas, HuggingFace, Git
Develop a NLU language model for patronising and condescending language (PCL) detection, submitted as part of the SemEval 2022 competition task. Work done include data analysis, cleaning and augmentation. Code utilised state-of-the-art RoBERTa model trained on the Don't Patronise Me! dataset.

3. Machine Learning 1 - Neural Network

Coursework Grade: 99% | TECH: Python, PyTorch, Numpy, Pandas, Git
A deep learning Neural Network regression model, built from scratch, used to predict housing prices based on the Californian Housing Dataset.

4. Machine Learning 2 - Decision Trees

Coursework Grade: 100% | TECH: Python, Matplotlib, Git
Decision Tree classification model, built from scratch, that predicts a user's location based on WiFi signals.

Software Engineering

1. Software Engineering Project - FuelOpt

Coursework Grade: 83% | TECH: Python, Django, Dart, Flutter, AWS, SQL, MariaDB, Docker, Git
Watch demo video here: demo video
A mobile application used to assist drivers in planning and optimising their fuel-filling journeys in terms of cost, time, and fuel efficiency. This was 13-week long Software Engineering Group Project where I was responsible for developing the optimisation algorithm, maintaining the backend server and our APIs.

2. Distributed Algorithms - Banking Raft Algorithm

Coursework Grade: 95% | TECH: Functional Programming using Elixir
Implement and evaluate a simple replicated banking service using the Raft consensus algorithm.

3. Blockchain - Tic Tac Toe

Coursework Grade: 100% | TECH: Solidity
Implement a tic-tac-toe game using Solidity.

4. Image Processing and Workflow

Coursework Grade: 90% | TECH: Python, Azure
Use different scheduling algorithms to assign and assess image processing workflow on Azure servers. This was done as part of the Scheduling and Resource Allocation course.