About
Hi, I’m Asrafil. I build production AI and enterprise software from Jakarta.
Full-stack and AI engineer, working remotely with teams in Indonesia and abroad since 2019.
Background
I started freelancing in 2019 through Pixelcraft, building client systems for teams in Indonesia and abroad. That work grew into a crypto payment gateway, a National Police IoT integration, an end-to-end mining data system for a publicly listed Indonesian company, more than 20 client projects, and Upwork Top Rated status.
In 2024 I joined PT Nuansa Cerah Informasi in Bandung, delivering four concurrent projects across Laravel, ASP.NET, and Next.js while introducing standard Git workflows, Docker, and CI/CD pipelines.
Since August 2024 I’ve been a fullstack engineer at ADW. I set the Next.js and Node.js baseline for an enterprise procurement system now used by 10+ engineers, and I build the internal MCP integrations that take repetitive work off the team.
Most of my work now sits where AI meets existing systems: retrieval over real documents, agents that call real APIs, and the evaluation and review steps that make both safe to run in production. I write about it on the blog.
How I work
In my own words, from the writing
-
Evaluate on every deploy
“RAG systems degrade silently after a model update, an embedding library bump, or a corpus change. Automated evals catch it before your users do.”
RAG pipelines -
Ground every answer
“With legal documents, hallucination isn’t a UX problem. It’s a liability problem.”
RAG pipelines -
A human approves the writes
“Human-in-the-loop checkpoints for any write operation. The model drafts; a human approves before create_approval_request fires.”
MCP integrations -
The server is the boundary
“Never expose raw database access through MCP. Your server is the security boundary.”
MCP integrations -
Test the whole path
“The end-to-end synthetic test is the most valuable. It’s the only metric that catches cascading failures across multiple layers simultaneously.”
IoT at scale -
Keep infrastructure boring
“Everything is containerised with Docker and deployed on a single beefy VM. No Kubernetes until you need it.”
RAG pipelines
Currently
- Role
- Fullstack Engineer at ADW
- Availability
- Open to work: remote, full-time or freelance
- Based in
- Jakarta, Indonesia (UTC+7)
- Focus
- RAG, MCP, agent workflows, and enterprise software
Elsewhere
- [email protected]
- GitHub
- asrafilll
- asrafilr
- RSS
- Writing feed