My Time at NUS, Singapore

Originally published on HackerNoon

Singapore hosts premier Computer Science institutions with unparalleled cutting-edge research. Both Nanyang Technological University (NTU) and the National University of Singapore (NUS) maintain worldwide reputations for exceptional CS programs, particularly in Artificial Intelligence.

The Opportunity

As a student at SRMIST in Chennai, India, I participated in the Global Academic Internship Programme (GAIP) by Corporate Gurukul. This initiative sends AI and Big Data-focused students to study under world-class faculty at NUS. The 15-day internship in December 2018 required completing two major projects.

Artificial Neural Networks Module (8 Days)

We worked under three incredible instructors: Dr. Lek Hsiang Hui, Dr. Tan Wee Kek, and Dr. Wang Wei.

Data Analytics & Decision Models

Dr. Lek introduced us to data analytics, decision models, and the full data pipeline — from extraction and cleaning to aggregation, representation, and interpretation. We picked up R programming fundamentals and dug into statistical measures covering location, shape, dispersion, and association.

Machine Learning Foundations

Dr. Tan covered the ML fundamentals: linear regression (simple and multivariate), classification methods (decision trees, Bayesian classifiers, logistic regression, SVM), clustering (K-Means, K-Medoids, hierarchical), and text mining. We worked extensively with Python libraries — NumPy, SciPy, Matplotlib, and Scikit-Learn.

Deep Learning

Dr. Wang took us deep into neural networks — from logistic regression limitations to gradient descent algorithms and backpropagation. We covered advanced optimizers (SGD, Minibatch GD, RMSProp, Adam), training techniques (random initialization, ReLU, dropout), CNNs (pooling, padding, strides), and RNNs including LSTM.

The Quick Draw Project

Our team built a real-time drawing prediction system using Google's Quick Draw dataset (4 classes, 20,000 images each). We tested multiple approaches — SVM, K-Means, feed-forward neural networks, CNNs, and LSTMs. CNN and LSTM achieved the highest accuracy. The frontend used OpenCV to capture keyboard stroke inputs in real-time.

Big Data Module

Under the legendary Ravindra Kumar (affectionately called the "God of Linux"), we dove into the Hadoop ecosystem: HDFS, YARN, MapReduce, HBase, ZooKeeper, Pig, and Hive. We configured a three-node Hadoop cluster deployed on Ambari, set up passwordless SSH, configured Java/JDBC/Hadoop environments, handled HBase data transfers, and ran MapReduce operations on text from Project Gutenberg.

Takeaway

This was a really enriching experience — learning from world-class faculty while shipping real projects in just 15 days. Huge thanks to Teaching Assistants Puru Sharma and Devvrit Khattri for their continuous support throughout the program.