EXPERIENCE
AI Engineer (Contract)JUL 2025 – JAN 2026
INDOCHINA HERITAGE TRAVEL · SINGAPORE
- Worked closely with the sales team to implement an LLM + n8n agentic workflow that converts Zoom transcripts into structured client proposals, eliminating manual write-ups and cutting prep time by 60%, contributing to $200K in closed deals.
- Automated the inbound email system by building a sentiment analysis pipeline with PII masking that auto-generated branded XeTeX PDF reports and notified team members through WhatsApp for record keeping and conversation tracking.
Research Assistant (PT)JUN 2025 – DEC 2025
NIE · SINGAPORE
- Collaborated with education researchers and teachers at NIE to build a RAG analytics platform that automatically surfaced utterance-level breakdowns across 100+ lectures, replacing hours of manual transcript review and adopted by 20+ teachers before I left.
- Improved retrieval accuracy by implementing multi-tier RAG with agent routing aligned to Singapore Teaching Practice frameworks.
- Extended the platform with a sandboxed Matplotlib environment so the chatbot could generate inline utterance distribution graphs on demand.
AI Engineer InternJAN 2024 – MAY 2024
PANASONIC R&D · SINGAPORE
- Cooperated with the HVAC R&D team to address a shortage of real energy data for model training by building WGAN, WCGAN, and DCGAN models on AWS SageMaker, generating synthetic time-series datasets that achieved 80% similarity to real distributions across 1-year intervals.
- Fed the synthetic datasets directly into downstream predictive optimisation models, reducing dependence on scarce real-world data.
RESEARCH
FYP · KAN-Enhanced GNNs2024–25
NTU · PROF. CHEN LIHUI
- Proposed replacing MLP activations in GNN architectures with KAN layers under Prof. Chen Lihui, designing GCNKAN and KANGraphSAGE that beat baseline GraphSAGE on all three benchmarks: 80.8% on Cora, 79.2% on PubMed, and 70.4% on CiteSeer.
- Demonstrated improved ROC-AUC and interpretability, though at the cost of 10x parameter overhead and increased overfitting susceptibility.
Fine-Grained Image Recognition2023–24
NTU · PROF. JIANG XUDONG
- Implemented an HSE framework on ResNet-50 with cascaded branches across 4 hierarchy levels under Prof. Jiang Xudong, achieving 81.3% top-1 on CUB200-2011 and 83.7% on Butterfly-200.
SELECTED PROJECTS
SEC Insider Trading Detector2025–
Automated SEC EDGAR Form 4 monitor with LLM-powered multi-factor scoring across insider history, transaction size, role seniority, earnings proximity, and market sentiment. 160+ positions tracked · ~15% avg PnL · +50% to +800% returns over 20–60d holds.
GrowthOS Knowledge MCP2025
LLM-wiki style FastMCP server that plugs into Claude to aggregate and persist learnings into user-specific Supabase buckets. Hybrid vector + keyword search across 500+ indexed items.
Slay the Semester2026
Cambridge x Incubed EduTech Hackathon · Slay the Spire-inspired gamified learning platform with deck building from completing weekly content and boss battles for assessments.
EDUCATION
M.S. Artificial Intelligence2026–27
UTS · SYDNEY
Adv NLP, Reinforcement Learning, Adv Data Analytics, Cloud & SaaS, SAS Predictive Analytics.
B.Eng. Electronic & Electrical Eng2021–25
NTU · SINGAPORE
Dean's List AY23–24, AY24–25 · ML & Data Specialisation.
DSP A+ · Image Processing & CV A+ · Pattern Recognition & DL A · ML Design A+ · Computational Intelligence A+.
SKILLS
LANGUAGES
PYTHONTYPESCRIPTSQLJAVASCRIPT
LLM / AI
LANGCHAINRAGAZURE OAIMCPPYTORCH
SYSTEMS
FASTAPINEXT.JSPOSTGRESDOCKERSUPABASE
ACHIEVEMENTS & LANGUAGES
Achievements
IdeasJam 2023 2nd Runner Up · RPA x GenAI Hackathon 2024 (SimplifyNext / NTU) 2nd Runner Up
Languages
Native English (IELTS 8.5) · Native Vietnamese · Conversational Mandarin Chinese