I build production AI systems that run businesses on autopilot — n8n workflows, multi-channel chatbots, ad operations, and Generative Engine Optimization.
Not demos. Production systems that handle real customers, real spend, and real revenue — built to be observed, owned, and modified.
Self-hosted n8n + Claude systems: multi-channel chatbots, cart / checkout / browse recovery, CRM sync, review pipelines. Zero vendor lock-in, ~$5/mo to run.
Meta & Google Ads recovery and optimization — diagnosing stuck campaigns, cutting wasted spend, fixing conversion tracking, and navigating policy restrictions.
Making brands visible to AI search. Schema architecture, llms.txt, FAQ markup, and citation-ready content so AI assistants actually cite you.
Real client engagements. Every number below is exact and copied straight from the work.
A Malaysian D2C brand drowning in 15+ daily support inquiries, fragmented across channels, with a static chatbot that broke on edge cases and 7,688 reviews scattered with no unified Shopify presence.
An Indonesian charity invisible to AI citations — broken schema markup, no structured data for the charity entity, generic meta descriptions, and no llms.txt.
A Singapore clinic's orthopedic campaign ran 0 impressions for 8 weeks despite "Eligible" status, while the ED campaign burned 97% of budget on irrelevant searches. The real cause: a Google classifier misclassification loop.
AI voice-cloning scams are exploding and existing tools (Hiya, RoboKiller) were built for the 2015 spam model. Built the full product solo — landing page, dashboard, voice infrastructure, billing, and scam database.
Clients need workflow logic that's observable and modifiable, not an opaque API-first black box. Designed a reusable 3-layer architecture — Triggers → Orchestration → Delivery — documented as a template.
I'm a solo founder building CheckTheCaller, an AI scam-defense SaaS — and I fund the build by running AI automation, ad operations, and GEO for clients across e-commerce, charity, and healthcare.
That mix is the point. I'm not a theorist who's read about AI workflows — I run them in production every day, for my own product and for clients. When I hand you a system, it's because I've already debugged the same class of problem at 2am on something I own.
My approach is metrics-driven and no-fluff: self-hosted infrastructure you actually own, observable logic you can modify, and exact numbers instead of vague promises.
Open to consulting and contract work in AI automation, ad operations, and GEO.