Projects
Portfolio case studies across LLM systems, machine learning, evaluation workflows, and decision-support products.
Projects
Flagship systems first, then supporting work by problem shape.
This page focuses on case studies across document intelligence, finance, forecasting, pricing, evaluation systems, and workforce intelligence. Most projects link to public code, articles, or demos; a small number are documented case studies where the source repo is private.
Each page stays short: business problem, outcome, key decisions, system design, and links to code or demos.
Projects are the canonical portfolio layer. Articles and demos are secondary surfaces linked from each project where available.
How to use the page
Start with the flagship set, then use the sections below for breadth.
The flagship set is the clearest signal for platform architecture, evaluation rigor, and operator-facing product work.
The supporting sections broaden the portfolio across private platform work, narrower NLP systems, decision tools, planning workflows, and applied ML studies.
The ordering is intentional: platform work first, then planning and decision systems, then the broader modeling range.
Cards marked Live demo, Article, or Repo have deeper public surfaces. A Private repo label means the case study is public but the source code is not.
Start here
Four flagship case studies that best represent the portfolio right now.
These are the strongest signals for platform architecture, evaluation discipline, and operator-facing product work.
Flagship case study
Creator AI Platform
Backend-first orchestration platform for AI-generated learning assets with quality gates, retrieval, and human review.
Flagship case study
Intelligent Content Analyzer
Document upload, retrieval, generation, and evaluation decomposed into a service-oriented product stack.
Flagship case study
SlideBench AI Evaluation Workbench
Evaluation workbench for AI-generated slides and learning artifacts using retrieval, provenance checks, and judge-assisted scoring.
Flagship case study
AI Portfolio Advisory System
Foundation-model risk profiling and objective-aware portfolio optimization in one advisory surface.
Supporting 3
Three supporting systems that add depth around current platform, evaluation, and knowledge-work themes.
This tier keeps the portfolio broad without diluting the flagship set: one private workforce system, one public LLM evaluation system, and one public enterprise RAG stack.
Supporting case study
Workforce Risk Intelligence Prototype
Public-signal intelligence workflow for workforce-risk monitoring, retrenchment forecasting, and operator dashboards.
Supporting case study
DSPy Automotive Extractor
Prompt optimization turned into a measured engineering loop with structured evaluation and local inference.
Supporting case study
Custom RAG Engine for Enterprise QA
Enterprise document QA built for privacy, local deployment, and mixed code-plus-text retrieval.
Supporting platform and language systems
Narrower NLP and language applications beyond the flagship and supporting sets.
Language systems
Earnings Report Intelligence Platform
Financial earnings report intelligence using traditional NLP and transformer-based analysis.
Language systems
YouTube Sentiment Analysis Platform
YouTube comment intelligence with transformer-powered sentiment analysis at production scale.
Decision systems and planning
Forecasting, optimization, and decision support tied to action beyond the top set.
Decision support
Dengue Forecasting and Intervention Economics
Time-series forecasting and health economics for intervention planning.
Decision support
Agentive Inventory Management System
Forecast-driven inventory planning with approvals, procurement guardrails, and auditability.
Decision support
Retail Price Optimization for CS Tay
Segmentation, price elasticity, and commercial strategy analysis for revenue optimization.
Applied ML studies
Modeling work that broadens the range of the portfolio further.
Applied ML
Elliptic Graph ML for Illicit Transaction Detection
Leakage-safe graph learning for illicit transaction detection with operational evaluation metrics.
Applied ML
ML Trading Strategist
Algorithmic trading across rule-based, tree-ensemble, and reinforcement learning approaches.
Applied ML
HDB Resale Price Predictor
Regression modeling for Singapore public housing price behavior.
Applied ML
Wet-Bulb Temperature Analysis for Singapore
Climate resilience analysis through wet-bulb temperature modeling and heat-stress interpretation.
Next step
Each project page links deeper into code, articles, and demos where available.
Use the project pages as the main portfolio layer. The live demos page remains a secondary index for browsing interactive surfaces only.
