Building AI-Powered
Systems That Scale

AI Compliance Review Platform

Manual compliance reviews were slow, error-prone and impossible to scale across hundreds of policy documents.

Built an LLM-powered platform that auto-reads documents, flags regulatory gaps and generates structured audit reports — cutting review time by 70%.

AI Spot Report Generator

Operations teams spent hours manually compiling data into reports, delaying decisions and wasting analyst time.

Delivered an LLM engine that ingests live operational data and auto-generates structured spot reports — reducing effort by 80%.

Document Chat Using AI

Employees wasted hours searching through thousands of documents to find answers buried in PDFs, Word files and reports.

Built a RAG-powered chat UI so users simply ask a question and get exact answers from the document library in seconds.

Agentic AI – Internal Ecosystem Bot

Repetitive internal workflows (approvals, lookups, handoffs) required constant human coordination across multiple systems.

Designed a multi-agent AI system that autonomously plans tasks, calls tools and completes cross-system workflows — freeing teams for high-value work.

AI Based Performance Optimization

Teams only discovered performance bottlenecks after users complained — reactive firefighting was costly and damaging to SLAs.

Built an ML monitoring platform that proactively detects degradation patterns and auto-recommends or applies fixes before users are impacted.

AI Sales & Lead Generator

Sales teams spent 60% of their time on unqualified leads and generic outreach, resulting in low conversion and wasted effort.

Engineered an AI pipeline that scores prospects, crafts personalised outreach via LLMs and feeds qualified leads directly into the CRM.

AI Customer Support Bot

Support teams were overwhelmed with repetitive queries, causing long wait times and high operational cost.

Deployed a RAG + LLM chatbot that handles 70% of queries autonomously, with full context awareness and 24/7 availability.

Predictive Analytics Engine

Business decisions were made on gut feeling — lack of forecasting led to inventory mismatches and missed revenue opportunities.

Built an Azure ML platform that forecasts demand, predicts churn and surfaces customer behaviour patterns in real-time Power BI dashboards.

Face Recognition Login

Password-based logins were a security liability — brute-force attacks, credential sharing and resets created constant risk and friction.

Delivered a facial recognition + liveness detection system for passwordless, phish-proof enterprise authentication.

AI Route Optimization System

DHL's manual route planning failed to account for real-time traffic and capacity — drivers followed suboptimal routes, increasing cost and delays.

Architected an AI system that ingests live data and dynamically re-optimises delivery routes — cutting fuel costs and improving on-time delivery.

Sentiment Analysis Dashboard

Customer feedback was scattered across emails, reviews and social media — no way to spot trends or act on dissatisfaction quickly.

Built an NLP pipeline that ingests multi-channel feedback and surfaces real-time sentiment trends in Power BI for immediate action.

Document Management System

Enterprise-grade multi-tenant DMS with role-based access, PDF viewing, email & WhatsApp sharing and audit trails.

URL Shortener & Analytics

Custom URL shortening service with click tracking, analytics dashboard, geographic insights and AdSense integration.

Automated Language Translator

Language barriers blocked effective communication between global teams and users across regions.

Designed an NLP-based automated translation engine supporting multi-language pairs with context-aware accuracy.

Image-to-Video & Text Classifier

Unstructured visual and textual content was impossible to categorise at scale without massive manual effort.

Built a deep learning pipeline that classifies images, converts image sequences to video summaries and categorises text content automatically.

AI Food Detector & Classifier

Training food recognition models from scratch required huge labelled datasets and significant compute resources.

Applied transfer learning on pre-trained CNN models to build a high-accuracy food detection and classification system with minimal training data.

Human Disease Predictor

Clinicians lacked tools to identify high-risk patients early — diseases were often caught too late from fragmented medical histories.

Developed an ML model that analyses patient medical history to predict disease likelihood, enabling early intervention and preventive care.

Blockchain & AI Integration

AI models lacked transparency and tamper-proof audit trails — enterprises couldn't trust or verify AI-driven decisions.

Investigated and implemented a framework combining blockchain's immutable ledger with AI decision logging to create verifiable, trustworthy AI pipelines.

Association Platform Framework

Hidden relationships between items, users and behaviours were invisible — businesses missed cross-sell and pattern opportunities.

Built an association analysis framework using graph-based algorithms to surface item relationships, behavioural patterns and recommendation signals.

Data Processing & Cleansing Pipeline

Dirty, inconsistent data across systems caused downstream analytics failures and unreliable business reports.

Built automated ETL pipelines with validation, deduplication and cleansing rules — ensuring analytics and ML models always run on trusted, verified data.

Automated Anomaly Detection

Fraudulent transactions, system failures and security breaches went unnoticed until significant damage had already occurred.

Developed an ML-driven anomaly detection system with continuous monitoring, auto-alerting and self-improving accuracy over time.

Point of Sale Software

Complete POS system with inventory management, barcode support, debt tracking, and detailed sales reporting.

Multilingual Voice Assistant

Frontline staff across three countries could not use the internal knowledge tools, which existed only in English.

Built a speech-to-speech assistant handling English, Malay and German — transcribes, answers from the same RAG index and replies in the caller’s own language.

Invoice Extraction Engine

Finance keyed thousands of supplier invoices by hand each month, with typos surfacing only at reconciliation.

Delivered a vision + LLM pipeline that reads any invoice layout, extracts line items to structured JSON and flags mismatches before posting — 92% straight-through processing.

Adaptive Learning Platform

Every learner got the same fixed course, so strong students were bored and struggling ones quietly fell behind.

Built a recommendation engine that scores each answer, models the learner and re-sequences the next module in real time — lifting completion rates by 38%.

Real-Time IoT Telemetry Hub

Sensor data from thousands of field devices arrived in bursts that overwhelmed the legacy nightly batch job.

Architected an event-driven ingestion hub on Azure Event Hubs and Service Bus, streaming to time-series storage with live dashboards and sub-second alerting.

Contract Risk Analyser

Legal review of vendor contracts took days per document, and risky clauses were still missed under deadline pressure.

Built an LLM analyser that compares each clause against the approved playbook, scores deviation risk and drafts redline suggestions for the lawyer to accept or reject.

Delivery Portfolio Dashboard

Leadership had no single view of delivery health — status came from slide decks assembled by hand every fortnight.

Built a portfolio dashboard pulling live data from Azure DevOps and Git, surfacing throughput, risk and dependency clashes across six teams without anyone updating a spreadsheet.