Singapore · B2B AI & SaaS company

• Primary architect and lead developer of a multi-tenant SaaS platform for e-commerce pricing and promotions. Grew it to 1,000+ B2B customers across 70+ countries at sub-3% 90-day churn. Built webhook and pricing events pipeline handling multi-million events a day.
• Architected and built an Heroku-style, agent-first cloud hosting PaaS (Hostbento Cloud) for AI-assisted development workflows, with zero-downtime deployments, rollback, and observability. PaaS supports multiple cloud providers.
• Designed the agent orchestration harness, real-time client sync over WebSockets, and the full conversation to generation to hosting pipeline behind the company's agentic website builder and hosting, Hostbento. Over 10,000 websites built and hosted.
• Built a website analytics event-ingestion pipeline handling over a million events a day, with an analytics database tuned for sub-second dashboard queries.
• Built an internal agent-deployment platform with one-click deploys, context versioning, rollback, and observability for AI agents and assistants such as Hermes.
• Designed and ran a private bare-metal data centre, handling capacity, network, nginx load balancing, DR, and monitoring. Cut p99 latency from ~2s to under 100ms at roughly 2.3× lower cost than the previous GCP Cloud Run setup. Multi-node Postgres cluster with continuous replication and MinIO cluster for CDN.
• Gathered customer requirements, built AI-native custom solutions for businesses, and deployed them on the Hostbento Cloud platform.
• Set technical direction and review standards across the stack, maintained Konigle SDK, and mentored more than 10 engineers and interns.

Tech Stack: Python, Django, Vue.js, Celery, Postgres, ClickHouse, MinIO, Github Actions, Nginx, FastAPI, WebSockets, Redis

Singapore · Entrepreneur First (pre-seed)

• Built on-device deep learning computer-vision models for edge hardware using quantization, pruning, and custom OpenCL kernels. Reached ~12–15× inference speedup with minimal accuracy loss.
• Shipped an end-to-end IoT stack into production pilots: edge device management, a real-time video-analytics pipeline, and a cloud backend (Node.js, React, WebSockets, MQTT, TimescaleDB).
• Led technical discussions with customers and partners, and deployed pilot systems into customer environments.

Tech Stack: C, C++, Node.js, React, TensorFlow, WebSockets, MQTT, TimescaleDB, Deep Learning, Object Detection

Singapore

• Optimized computer-vision and deep-learning models for real-time performance in shipping consumer-electronics products. Built a deep learning inference library. Mentored junior engineers on model optimization.

Tech Stack: C, C++, OpenCL, Deep Learning, Object Detection

Singapore

• Published award-winning research on efficient deep-learning inference for constrained hardware, including the ESWEEK 2016 Best Paper. Authored CaffePresso, an open-source CNN-acceleration library for embedded SoCs.