Operating inventory
A map of public tools, demos, and surfaces I use to make the work legible. The common thread is applied AI, product judgment, systems engineering, and enough design taste to ship the whole thing.
Make the operating loop visible.
Prefer a working surface over a static explanation.
Use design taste as engineering leverage.
Keep public proof useful and private work private.
Inventory
A public model for inspecting applied AI agents: context freshness, decision quality, guardrails, human review, and learning loops.
Connects the product, data science, and engineering questions that make AI transformation inspectable instead of theatrical.
A concise operating note for applied AI work shaped by Singapore and APAC product contexts, including AAmplify, FlumeWorks, agent infrastructure, and production decision loops.
Designed to make the public thesis discoverable without turning the homepage into a keyword page.
Themes, articles, tags, and Cmd+K search for organizing field notes, TILs, decision science, realtime systems, and agentic work.
Built around overlapping tags and broader buckets so writing can support both readers and a more disciplined drafting workflow.
A role-by-role record of scaled systems: core transport booking, OMS/order state, payments, personalization, high-volume communications across push/SMS/email and owned in-app product surfaces, finance systems, and realtime learning.
Shows the veteran engineer arc behind the current applied AI and product-building work.
Access-controlled workspaces for client or partner explorations where the useful artifact is a working surface, not a static deck.
Keeps sensitive material private while preserving the habit of shipping end-to-end prototypes.
A small artillery game that makes Thompson Sampling tangible: the AI explores, exploits, adapts to changing wind and terrain, and relearns when static rules break.
Built to help people feel why adaptive algorithms beat fixed rules in changing contexts, and to make the Aampe-style decision loop easier to understand in a toy world.
How this fits
The site is becoming an operating system for the work: writing turns field lessons into reusable thought, Cmd+K makes everything fast to retrieve, and the public build inventory gives readers a practical route into the actual surfaces.
That is also the SEO, AEO, and GEO strategy: be findable for applied AI and AI transformation in Singapore by showing the work, the operating model, and the engineering judgment behind it.