Founding CTO
Solo-built B2B platform that gates contractor disbursements on VLM-verified photo evidence across thousands of housing units nationwide.
AutoKon is a B2B financial verification pipeline for Indonesian property developers. Photo evidence from construction sites gates contractor disbursement — a tamper-resistant record that work was done before money moves.
The platform demonstrates end-to-end ownership of a regulated financial workflow: multimodal AI, agentic interfaces, eval discipline, and multi-instance orchestration — shipped solo at production scale.
Construction operations needed faster coordination, auditable progress evidence, and payment safety — but process data lived in disconnected systems, manual handoffs delayed disbursements, and unverified AI descriptions could misroute contractor payments.
Shipped a full-stack production platform solo (Next.js, Supabase, TypeScript) with S-curve / earned-value dashboards and a disbursement engine. Deployed VLM image analysis (Claude Opus Vision + Pinecone), an LLM-as-judge evaluation pipeline before AI output gates money, WhatsApp-first field entry via n8n + Botpress, and a 5-role photo-evidence approval workflow.