Danson

Full-stack engineer based in Singapore with 4+ years of experience building scalable backend systems and the frontends that sit on top of them. Currently at Vistra Singapore; previously at DCS Card Center, where I worked across TypeScript, Spring Boot, and PostgreSQL — focused on API performance, observability, and turning slow, manual workflows into fast, automated ones.

I care about the parts of software that compound: clean queries, good monitoring, type-safe interfaces, and code that the next engineer can actually read.

I do most of my development with AI agents: Claude Code day to day, and Amazon Bedrock at work. AI writes the first draft; tests, review and my own checks decide what ships.

• Tests first. I write the tests from the spec before the agent writes any code, so "done" means the tests pass, not that the agent says so. AI also reviews my diffs as a second pair of eyes.
• Grounded, not guessed. My agents are set up to ask when something is unclear and to back answers with docs and research, so I learn the context instead of trusting the output blindly. Hooks compact long sessions before the context grows large enough to cause drift, saving a checkpoint first.
• Cost-aware. Different agents run on different models: cheap ones for lookups and boilerplate, the strongest only for changes that span several parts of a system. A terse output mode cuts token use further.
• Internal tools at work. At Vistra I build internal tooling on Amazon Bedrock, including custom MCP servers that give coding agents access to our codebase.
• Verified in production. A deploy once reported success while the server kept running the old config. I caught it by checking the live site, not the agent's summary.
• Agents as pipeline stages. For a side project making short Singapore history videos, Claude Code writes scripts from my research notes and picks archive search terms; Node scripts fetch licensed images, render the video and add the voiceover. I approve every script and image by hand.