M.Sc. Data Science in CCDS

Nanyang Technological University, Singapore

"After 5 years at POI-TECH — leading teams, earning recognition, and operating in my comfort zone — I chose to step away. Not because I had to, but because I knew AI was reshaping everything, and I wanted to be on the right side of that shift. Singapore, CCDS, and a deep dive into data science felt like the only honest next step."

Academic Projects

PythonRaBitQPCAK-Means

Mar – Apr 2026

High-Performance Vector Search System on DBLP Dataset via RaBitQ

100k+ docs

Processed via byte-stream parsing with 384-d semantic embeddings

1-bit quantization

RaBitQ + Popcount — memory reduced to tens of MBs

66% compute cut

PCA 384→128 dims while preserving recall > 90% variance

High-Performance Vector Search System on DBLP Dataset via RaBitQ diagram

◆Processed 100k+ documents via byte-stream parsing and generated 384-d semantic embeddings.
◆Designed IVF + K-Means index to reduce search space and improve query efficiency.
◆Applied RaBitQ quantization (1-bit) with Popcount acceleration, reducing memory to tens of MBs.
◆Optimized with PCA (384→128), cutting compute by 66% while preserving recall >90% variance.

PythonLLM EvaluationData Analysis

Mar – Apr 2026 (ongoing)

Safety Evaluation of Generative Autonomous Driving Videos

100+ videos

Labeled dataset for driving video safety evaluation

Multi-level errors

Semantic, logical, and decision-level failure detection

LLM-powered

Automated prompts for safety scoring at scale

◆Built a safety evaluation pipeline for generative driving video models, detecting semantic, logical, and decision-level errors.
◆Created a labeled dataset of 100+ videos and developed LLM-based prompts for automated safety evaluation.
◆Conducted failure mode analysis to quantify and analyze safety issue distributions.

PostgreSQLPython

Mar – Apr 2026

Large-Scale DBLP Database System: Parsing, Modeling and Optimization

PostgreSQL

Robust relational schema for academic publication data

XML → SQL

Full pipeline from raw DBLP XML to structured tables

Indexing study

Scalability experiments across different DB sizes

◆Architected a robust PostgreSQL relational database for large-scale academic publication data.
◆Parsed and transformed the DBLP XML dataset into structured tables for analysis.
◆Evaluated query performance and scalability by experimenting with different database sizes and indexing strategies.