Alif Daffa’ Yusof

👨🏻‍💻 About Me

💼 Work Experiences

A*STAR (Institute for Infocomm Research)

• Explored methods for facial expression recognition from video data, including various feature extraction methods like Facial Landmark extraction, Histogram of Oriented Gradients, and OpenFace toolkit, as well as models like Recurrent CNNs, LSTMs and Gated Transformers.

• Co-authored a paper on above work for the IEEE International Conference on Data Mining (ICDM) 2023 workshop.

• Spearheaded the research and development of a self-supervised federated learning implementation for dynamic facial expression recognition.

• Published a paper to the International Conference of Computers in Education (ICCE) 2023 workshop as first author.

Multi-Agent Robot Vision & Learning (MARVL) Lab

• Conducted literature reviews & experiments for evaluating deep learning based visual odometry methods.

• Co-authored a research paper submission titled Evaluating Visual Odometry Methods for Autonomous Driving in Rain.

• Implemented various Conditional Generative Adversarial Networks using PyTorch to remove rain droplets from car camera footage, allowing for improved accuracy of vision-based localization and perception algorithms during adverse weather.

• Developed scripts for training, evaluating, and performing inference with various image de-raining models and datasets.

Collins Aerospace

• Merged and cleaned operational data from multiple tables using MySQL to curate a suitable machine learning dataset.

• Trained an XGBoost model to predict the risk of an equipment’s repair being late and deployed it to a Flask server.

• Optimized above model using feature engineering and Optuna hyperparameter tuning, raising cross-validation accuracy by 20%.

• Developed an interactive web dashboard using Dash Plotly where the above model is used to improve servicing job prioritization.

• Integrated dashboard with a SQL database for warehousing and tracking status history of jobs, saving 20 hours of labor monthly.

Temasek Labs @ SUTD

• Developed a dynamic simulation environment in Unity Game Engine for testing of intelligent human – multi-robot teaming.

• Co-authored a paper to the peer-reviewed IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR) 2022.

🛠️ Featured Projects

Music Captioning Project

This project was developed as part of the 50.038 Computational Data Science module.

• Led a team of 3 in developing Music Captioning models on Google’s MusicCaps dataset.

• Evaluated the performance of various pre-trained audio feature extractors such as MusicNN, PANN & Wav2Vec2 in an encoder-decoder architecture, demonstrating how relevance of pre-training tasks directly impacts model performance in music captioning.

Tags: Deep Learning NLP

Chest X-Ray Multi-Label Classification

This project was developed as part of the 50.039 Deep Learning module.

• Worked in a team of 3 in developing a model to identify thoracic diseases from chest X-ray images.

• Trained and evaluated advanced model architectures such as ResNet, DenseNet & Vision Transformers.

Tags: Deep Learning Computer Vision

 

MunchMatch

This project was developed as part of the National AI Student Challenge 2022. MunchMatch is a website which aims to allow Singaporeans to select and choose their ideal restaurant with ease and confidence.

• Developed an aesthetic & user-friendly website for users to search, filter and compare restaurants based on their cravings.

• Web scraped >40,000 restaurant reviews from 400+ restaurants on TripAdvisor using Selenium.

• Extracted specialties of each restaurant by designing a pipeline consisting of TF-IDF keyword extraction and aspect-based sentiment analysis using AI Bricks’ SenticGCN model.

• Implemented a semantic search engine for users to find relevant restaurants based on search queries.

Tags: Software Engineering Deep Learning NLP

American Express Default Prediction Competition

This project was part of the Default Prediction Kaggle competition hosted by American Express.

• Performed data preprocessing and feature engineering on 50GB of time-series data.

• Trained an ensemble comprising of a Recurrent Neural Network, XGBoost, and CatBoost models.

• Achieved a final ranking of 399th place out of 4874 participants worldwide (Top 9%).

Tags: Machine Learning Deep Learning

 

Hate Speech Classification

This project was done as part of the 50.007 Machine Learning course module.

• Led a team of 4 in testing & tuning various machine learning models such as Linear Regression, SVM & Random Forests.

• Experimented with data augmentation techniques such as SMOTE as well as Principal Component Analysis on TF-IDF features.

• Optimized cross-validation f1-score with various ensemble methods such as stacking, blending, and voting.

• Team’s submission placed as 2nd runner up among cohort mates in hidden test set as well as public test set.

Tags: Machine Learning

Singapore Autonomous Underwater Vehicle Challenge

This project was done as part of the Singapore Autonomous Underwater Vehicle Challenge, and also an Undergraduate Research Opportunity Program (UROP) project.

• Led a team of 6 in implementing autonomous navigation logic for an underwater robot using ROS.

• Trained an Object Detection model on a synthetic dataset generated using Unity’s Perception Toolkit.

• Implemented navigation logic using state machines, pathfinding algorithms, and custom maneuvers.

Tags: Software Engineering Computer Vision Deep Learning

Ascenda Loyalty Hotel Booking Platform

This project was developed as part of 50.003 Elements of Software Construction module.

The Ascenda Loyalty Hotel Booking Platform is a full-stack web development industry project, which made use of Ascenda's API endpoints to deliver a white-label hotel booking platform.

• Developed landing page, frontend components and routing between pages using ReactJS and CSS.

• Developed automated end-to-end system tests using Selenium.

Tags: Software Engineering

Teacher’s Space Android App

This project was developed as part of the 50.001 Information Systems & Programming module.

Teacher’s Space is an Android app to allow for messaging, calling, and scheduling between teachers and students.

• Developed the contact list & chat messaging feature using Firebase and Android Studio.

Tags: Software Engineering

Drone Simulation + AI

This was a personal project which combined my interest in AI with my experience in using Unity Game Engine.

• Developed an environment in Unity Game Engine for simulating a physics-based drone controlled by individual propellers.

• Trained a Proximal Policy Optimization (PPO) agent to pilot the drone to designated points in 3D space using Unity Engine’s ML-Agents toolkit and curriculum learning.

Tags: Software Engineering Deep Learning

Asphalt AI

This was a personal project which marked my first exposure to deep learning.

• Developed and trained a Convolutional Neural Network which plays Asphalt via real-time classification of screen images using behavioral cloning and is able to consistently outperform in-game bots.

Tags: Computer Vision Deep Learning

 

Text Based Pacman Clone

This project was developed as part of the 10.014 Computational Thinking for Design module.

• Created an engaging gameplay experience directly in the terminal, complete with characters, ghosts, points, and power-ups.

• Leveraged OOP concepts to create various game entities (Player, Ghosts, Tiles).

• Developed robust classes and methods for handling game logic, such as movements, collisions, and scoring.

• Designed basic AI for ghost characters, which allowed them to move around the map randomly and spawn at predetermined intervals. The ghosts also react differently based on the power-up status of the player.

• Integrated real-time keyboard input for character movement using the msvcrt library in Python.

Tags: Software Engineering

 

NamelessRPG

This was a personal project I developed prior to enrolling in SUTD as a fun way of gaining software development experience.

The game begins as an idle-incremental resource gathering game similar to A Dark Room, where new features and resources are gradually unlocked. Eventually, the player unlocks the option of training and equipping a party of villagers to explore the world, where the party will have to engage hordes of enemies in active time-based battles.

• Developed a playable game with numerous features using Unity Game Engine.

• Employed an event-driven programming approach to control game flow and allow for new features to unlock over time.

• Adhered to SOLID design principles to ensure the codebase was modular, flexible, and maintainable, thereby facilitating future updates and expansions.

• Managed all aspects of game development including game design, coding, debugging, and testing.

Tags: Software Engineering

 

📚 Education

⚙️ Technical Skills

Software Development Tools & Frameworks

Data Science Libraries

Programming Languages

Cloud & Big Data

Others

🏆 Achievements & Certifications

ACHIEVEMENTS

• American Express Default Prediction Competition – Ranked 399/4874 (Top 9%)

• Machine Learning Course Project (Hate Speech Classification) – Ranked 3/33 (2nd Runner Up)

• 30 Days of ML (Housing Price Prediction Competition) – Ranked 809/7573 (Top 11%)

CERTIFICATIONS