How I Work
I build production AI systems that turn hour-long banking workflows into minutes, engineered for regulated enterprise environments. As a practitioner shipping end-to-end applications for a large Indonesian state-owned bank, I partner with teams on document automation, grounded RAG assistants, and verification engines.
Engineering Methodology
Isolate the Operational Bottleneck
Identify where manual workflows fail—whether transcription delays, helpdesk queues, or tedious line-by-line paperwork comparison.
Pair extraction with human verification
Build specialized extraction pipelines paired with human-in-the-loop verification UX rather than relying on end-to-end black-box generation.
Evaluate with Multi-Run Suites
Test candidates against evaluation sets, at least three runs per question, before promoting them.
Stage, Audit, and Guardrail
Deploy to a central platform with staging, one-click rollback, and full audit trails.
Working with AI Agents
AI coding agents can work safely in a shared codebase when they have written standards, memory, and permission limits. Claude Code, MCP, and multi-agent workflows are part of my daily way of working.
Interested in shipping similar systems?
I partner with teams to design, evaluate, and ship production-grade AI systems with verified operational impact.