Year 3 Data Science & Analytics student at NUS building explainable AI systems, LLM-powered agents, and production-grade ML applications.
I focus on:
- Agentic AI workflows
- Retrieval-Augmented Generation (RAG) systems
- Hybrid ML + LLM architectures
- Explainable company intelligence systems
Currently exploring AI systems design, evaluation frameworks, and scalable AI product deployment.
- 🌱 Pursuing a B.Sc. (Hons) in Data Science & Analytics
- 💡 Strong interests in AI/ML, AI Agents, Machine Learning, and Product Development
- 🚀 Experienced in data science, operations, and leadership through internships and student organisations
- 🏆 Finalist at NUS SDS Datathon 2026 (Top 1/76 teams)
- 🏆 Finalist at NUS SDS Hackathon 2025 (Top 3/40 teams)
- 🏆 Finalist at NUS SDS Datathon 2025 (Top 5/75 teams)
- ✨ Believer in teamwork, adaptability, and continuous learning
- Multi-agent orchestration frameworks
- Vector database optimisation strategies
- Embedding evaluation techniques
- LLM cost-latency tradeoffs
- AI evaluation benchmarks (precision, hallucination detection)
- AI product management frameworks
-
Artificial Intelligence Entrepreneur Intern @ Crayon Data (Dec 2025 – Jan 2026) → Built LLM-powered data extraction and agentic workflows to transform unstructured merchant offers into schema-compliant datasets, reducing manual processing and improving data quality at scale.
-
Operations (Data Science) Intern @ Superbank (May – Aug 2025) → Engineered 200+ fraud detection features in Python/SQL, optimised queries on 50M+ records (~40% faster), built real-time monitoring pipelines in Snowflake.
-
Head of Branding & Marketing @ Developer Group, NUS Computing (Jul 2025 – Present) → Leading branding initiatives and managing associates for impactful events.
-
Curriculum Executive @ NUS Product Club (Jul 2025 – Present) → Organised flagship Product Management Executive (PME) programme with industry PMs.
-
Marketing Head @ NUS Science Club (2024–25) → Secured sponsorships, co-led 7 executives, and streamlined sponsor email processes (50% faster).
Explainable AI-powered company intelligence system combining classical ML clustering with LLM-based reasoning — fully grounded, no hallucination.
Traditional company intelligence tools lack explainability and rely on black-box scoring systems.
Built a hybrid ML + LLM architecture:
- K-Prototypes clustering on 8,559 companies
- 20+ engineered financial and IT intensity features
- Structured RAG-style retrieval (15 records/query)
- Guardrail-constrained Llama 3.3 70B via Groq
- Temperature-controlled outputs (0.1)
- Dockerised full-stack system
- Mixed-data clustering (numeric + categorical)
- Silhouette-based K selection
- Dynamic knowledge base generation
- Strict dataset-only grounding
- No hallucination policy enforcement
- React + FastAPI production pipeline
Demonstrates ability to:
- Combine classical ML with modern LLMs
- Design explainable AI systems
- Build safe and grounded AI workflows
- Ship end-to-end AI applications
- Built LLM-powered workflows to extract structured insights from unstructured merchant datasets
- Designed prompt pipelines for schema-compliant data generation
- Implemented guardrails for output consistency
- Worked with production-grade AI deployment workflows
- Worked on AI-driven data pipelines
- Applied ML models in regulated public sector environments
- Focused on accuracy, auditability, and stakeholder trust
- 🏅 NUS Datathon 2026 Finalist (Top 1/76 teams)
- 📚 Principal's Honours Roll (2019 & 2020)
- 🎓 Top cohort results for O-Level Mathematics & Science
- ⭐ Outstanding academic performance in JC1 & JC2
When building AI systems, I prioritise:
- Grounding over hallucination
- Classical ML + LLM hybrid architectures
- Explainability before automation
- Low-temperature deterministic outputs for business settings
- Clear guardrails and retrieval constraints
- Containerised deployment-ready systems
I believe production AI must be: Safe · Interpretable · Scalable · Cost-aware
Reach out to me via the platforms below! My resume is available on request.
