I'm Keith Sng

A fully automated, end-to-end ETL pipeline emulating production data ingestion and analytics engineering using Apache Airflow. Daily stock prices are pulled from the Yahoo Finance API, stored in PostgreSQL, and transformed with SQL to compute moving averages for analysis. Results are surfaced through an interactive Streamlit dashboard, enabling end-users to explore trends, analyse performance, and consume insights via data report downloads.

A product analytics project evaluating the feasibility and performance of Wise's new INR → USD currency route (hypothetical). Combined pre-launch demand estimation methods to assess market potential and profitability with post-launch funnel analysis across the first eight weeks to identify adoption patterns, conversion drivers, and friction points. Used Python and Pandas for data preparation, DuckDB for analytical querying, and Plotly for interactive visualisations, delivering actionable insights to inform product optimisation and strategic decision-making.

A marketing analytics tool that enriches website data by querying the WhatCMS API to identify underlying technologies such as CMS platforms, programming languages, databases, and web servers. Built for B2B marketing workflows, the application supports batch URL processing, rate limiting, and CSV & Excel input–output, delivered via a command-line interface (CLI) for lead qualification, segmentation, and market research.

A statistical analysis project evaluating the effectiveness of marketing campaigns through A/B testing. Uses t-tests, ANOVA, and regression models to assess whether campaign variants drive increases in product views and sales revenue, and identifies the level of effort required to maximise ROI. Demonstrates end-to-end experimental design, hypothesis testing, and insight communication.

A collection of machine learning projects covering classification, regression, clustering, and time-series forecasting. Includes end-to-end pipelines with EDA, feature engineering, hyperparameter tuning, and model evaluation. Projects span Titanic survivorship prediction, King County housing price regression, clustering of UCI wine datasets, and 60-day stock price forecasting using SARIMA and Holt-Winters methods.

Comprehensive analysis of Singapore’s public housing market using multiple HDB datasets (BTO prices, resale transactions, rental trends, housing grants). Performed data cleaning, feature engineering, and visualisation with NumPy and Matplotlib, generating insights on price patterns, estate affordability, financial support accessibility, and spatial housing trends across Singapore.

Comprehensive analysis of Singapore’s labour market using MOM, OECD, and World Bank datasets. Built a full ETL workflow to retrieve and clean data, stored it in a MySQL database, and conducted multi-dimensional analysis across unemployment, retrenchment, turnover, wages, working hours, leave benefits, and flexible work arrangements. Visualisations and PCA were used to uncover insights on job security, sectoral opportunities, and work–life balance for job-seekers in a tight labour market.