Personal Projects

---

Real-time Facial Emotion Recognition

A real-time facial emotion recognition app which uses your laptop's camera feed to detect faces and then classifies the facial expression (eg. happy, sad, surprised, angry, etc.). A convolutional network model was built in Python using Keras/Tensorflow and then deployed using OpenCV in C++.

Tech: Keras, Tensorflow, OpenCV, Python, C++

---

Quora Insincere Questions Text Classification (Kaggle)

Using NLP methods to detect which questions posed on Quora were insincere. I started off using a simple model baseline using bag of words and TFIDF features, and then improved upon my results using Deep Learning models (LSTM, Attention).

Tech: Python, NLP, Keras, Tensorflow

---

Visualising Bayesian Networks using Shiny and VisNetwork in R

I worked on a project with a large Australian Telco client where my team delivered a cutting-edge solution to quantify ROI of various marketing activities for a number of target variables, using a Bayesian Network methodology. I built a network visualisation using Shiny to present the results of the model to non-technical stakeholders. Note: Client details have been scrubbed and spend data has been replaced with random values.

Tech: R, Shiny, Bayesian Networks

---

IEEE-CIS Fraudulent Transactions (Kaggle)

Building models to detect fraudulent card transactions. Features in the dataset include the transaction time and amount, product code, card type, geography, email domain, device and a number of pre-built features by the competition hosts.

Tech: Python, Xgboost/LightGBM

---

NRL DataJam 2019

NRL DataJam is a Rugby League themed data hackathon which I entered as part of an Accenture team. Our team tackled the question of whether to opt for the penalty goal (2pts) or to attack for the try (up to 6pts). We incorporated a machine learning model and a Tableau dashboard that would guide a coach's decision for various scenarios. Our team was awarded 3rd place for our solution.

Tech: R, RandomForest, Tableau

---

Real Estate Valuation Dataset Creation by Web Scraping

(Work in progress) I built my own real estate dataset to help inform my personal property investment decisions and to upskill in web scraping. The project involves web scraping data from realestate.com.au and other property sites, and then in the future, analysing this data and building a model to estimate property value, given property size, location, bedrooms, etc.

Tech: Python, Selenium, Beautiful Soup