Experienced Full Stack Engineer skilled in cloud infrastructure, automation, and DevOps. Currently building AI evaluation and testing infrastructure at Resaro AI. Passionate about building scalable, secure, and cost-efficient systems that drive engineering excellence.
A builder and solution-focused architect who enjoys turning ambiguous problems into impactful, production-ready solutions. My work blends product thinking with strong fundamentals in systems and frontend engineering, supported by deep expertise in cloud-native technologies and DevOps practices.
I focus on enabling teams, improving developer experience, and driving technical excellence across scalable, high-performing systems — while continuing to write about engineering, mentor others, and explore new developer tooling.
Building Resaro's Approved Intelligence Platform — closing the gap most organisations face in testing and evaluating AI. Rather than shipping a report or a checkbox, the platform tests AI in real operational environments and produces the evidence that actually gates deployment.
Leading regional solution architecture across ASEAN and GCR, driving Redis Enterprise adoption through strategic solutions, technical leadership, and automation to deliver scalable, high-performance systems.
Delivered national-scale digital identity systems at GovTech's NDI platform, spanning serverless infrastructure, cloud cost optimisation, security compliance, and engineering leadership across multiple high-impact Singpass products.
Coordinated security and infrastructure projects, streamlining vulnerability management through automation and Kubernetes deployment to enhance compliance and audit readiness.
Collaborated cross-functionally to align engineering and product teams, ensuring timely releases and building a Jira-integrated analytics portal for workload and project tracking.
A Claude Code configuration template for full-stack projects using Next.js and FastAPI. Drop this into your project and Claude gets safety rails, domain skills, quality-gate commands, and structured PR workflows out of the box.
A Next.js 15 + FastAPI web application for the 2026 FIFA World Cup that displays fixtures, live scores, group standings, and knockout bracket progression. It includes an AI-powered prediction engine that generates match outcome probabilities with explanations, backed by PostgreSQL for persistence and Redis for caching.
About Retail shelf annotation demo using RedisVL vector search and Gemini Flash to identify and label products from a rack image
Multi-modal vector search (text, image, audio) powered by Redis — BLIP captioning, Whisper transcription, sentence-transformers embeddings, sub-5ms search latency.
A Next.js + Leaflet demo showcasing Redis Geo search for nearby EV chargers. Shows how Redis outperforms generic filtering by combining GEOSEARCH with hash lookups for fast, low-latency geolocation queries, with filters, sorting, and a live interactive map UI.
A Python-based implementation of semantic search and intelligent caching using Redis Stack, vector embeddings, and a local LLM (Flan-T5). Demonstrates fast KNN search and cache reuse for similar queries to reduce LLM calls.
Provision Redis Cloud on AWS with Terraform — automates Redis Enterprise Cloud setup including subscription, database creation with RedisJSON, RedisSearch, TimeSeries, VPC peering, and optional S3 remote state management.
Stream AWS Load Balancer, CloudFront and WAF logs stored in S3 into CloudWatch Logs to use with CloudWatch Insights or metrics filters.
Implementation of an authorization service deployed as a serverless application on AWS Lambda, using JWT (ES256), Secret Manager, DynamoDB, and Middy middleware.
An API that converts text to human-like speech, generating audio files hosted on AWS S3 with presigned URL access. Built with Flask and the pyttsx3 library.
A solution leveraging human interaction with a native application and active sensors such as AI cameras with OCR technology to crowdsource bus demand information.
Collaborated with NEA and leveraged IoT to collect tray return data at hawker centres, gathering insights to improve future campaigns.
Collaborated with Fresh Fruit Labs Group to understand their social media presence. Data collection from Instagram, Facebook, Burpple and Google Reviews using Selenium, including sentiment, network, and competitor analysis.
Published on Medium
A Jira ticket used to be three sentences, because whoever wrote it was busy and trying to be concise. Now it’s five paragraphs. There’s a background section, a problem statement, the acceptance criteria.
Read →Performance appraisal season. You open the blank review form and you’re asked the same questions every cycle: what did you accomplish and what did you learn?
Read →When I started my career, I thought being a good engineer came down to output. Ship the most features. Clear the most tickets. Collect the certifications that would prove on paper that I knew what I was doing.
Read →When I started my career, I thought being a good engineer came down to output. Ship the most features. Clear the most tickets. Collect the certifications that would prove on paper that I knew what I was doing.
Read →