Hello! I'm a NUS Data Science and Analytics graudate, and I focus on retrieval systems, RAG pipelines, multi-agent workflows, and ML/DL.
- Retrieval and RAG: hybrid retrieval, reranking, citation constraints, abstention, evaluation sets, and source-grounded generation.
- Multi-agent systems: LangGraph orchestration, specialist agents, tool routing, local LLM workflows, and cited synthesis.
- Machine learning: customer segmentation, campaign analytics, model evaluation, and experiment packaging.
- Deep learning: PyTorch language models, LSTM/Transformer comparisons, ablations, and generation quality checks.
| Project | What to look at | Stack |
|---|---|---|
pdpa-qa-system |
Citation-constrained RAG for Singapore's PDPA, with hybrid retrieval, reranking, abstention, and PDPABench evaluation. | Python, FastAPI, BM25, dense retrieval, cross-encoders |
AlphaAgents |
Multi-agent equity research system that routes S&P 500 questions across news, SEC filings, peers, market context, and sentiment agents. | Python, LangGraph, Ollama, Chroma, SEC EDGAR |
personalised-bank-marketing-campaigns |
Traditional ML and analytics modules for bank campaign strategy, customer segmentation, ROI, retention, and engagement. | Python, Docker, scikit-learn, analytics pipelines |
phish-n-cheats |
Marketplace scam-awareness simulation with planted listings, seller chats, scoring, reports, and trust-and-safety analytics. | TypeScript, React, Express, LLM-backed simulation |
shakespeare_lstm_transformer |
Character-level language modelling study comparing LSTM and Transformer architectures on Shakespeare. | PyTorch, LSTM, Transformer, ablations |
memory-garden |
Local-first interface for reviewing and editing assistant memory files through a visual garden metaphor. | TypeScript, React, Vite, File System Access API |