Agent HQ · Lobby
Software Engineer and AI Engineer (Azure)
I build software and AI on Azure that survives contact with real users.
I am a software engineer and AI engineer based in Jakarta. Since February 2026 I design, build and operate Azure AI applications at Metrodata for clients in banking, automotive, mining and food manufacturing: realtime voice agents, document AI, AI security tooling and grounded RAG assistants. Before that I spent four years building backends and full-stack products for e-commerce, travel, SaaS and digital banking, from MNC and Jatis to Jenius.
L1
2019
Shipping software since 2019, full-time since 2022
800+
Pull requests merged in 2026
7,350
GitHub contributions in the last 365 days (agent-assisted, human-reviewed)
3
Microsoft certifications (AI-103, AI-901, AZ-900)
20
Competition placings, 7 of them first places (2018 to 2022)
10x
Faster voice-agent interruption, ~2 s to ~0.2 s (controlled A/B)
L3
Software engineering since 2019 and enterprise AI on Azure since 2026, presented side by side.
L3 · Software Wing · Hero
Insurance partner microservices and an in-house SFTP service inside a digital bank.
5 partners Insurance partners integrated, with about 30% fewer operational errors
Jenius (SMBC Indonesia) · Jun 2024 to Dec 2025
L3 · Software Wing · Hero
Auctions, resale, a wallet ledger and partner settlement for a live ticket marketplace.
61% Of all repository commits authored by me (162 of 267)
TEMMPAT · Nov 2025 to Feb 2026
L3 · Software Wing · Hero
Purchase orders, invoices, tax checks and approvals, shipped solo in about six weeks.
~6 weeks From first commit to a deployed portal, as the only engineer
Aug 2026 to Oct 2026
L3 · AI Wing · Hero
Callers interrupt the agent naturally (~2 s to ~0.2 s), get grounded answers and reach a human mid-call.
~2 s to ~0.2 s Caller interruption during knowledge lookups
May 2026 to Oct 2026
L3 · AI Wing · Hero
Eight scanners on every pull request, AI triage people actually read, and a fix agent that opens guarded PRs.
8 engines Scanning engines on every pull request across a 100+ repository organization
Jun 2026 to Oct 2026
L3 · AI Wing · Hero
OCR plus a two-pass LLM review of procurement packets, with a risk engine that had to pass its own gates.
90.9% to 36.4% False-positive rate on noisy clean packets
May 2026 to Oct 2026
L2
Metrodata (PT Mitra Integrasi Informatika)
In the Microsoft AI practice I design, direct, review and ship Azure AI applications for clients in banking, automotive, mining and food manufacturing, and for the practice's own products, using spec-driven, AI-agent-orchestrated delivery.
Naremax
Backend modules for a digital products and training e-commerce platform, built with a 5+ person agency team in a TypeScript monorepo.
TEMMPAT
Led the v2 backend refactor of a premium event ticket marketplace into a modular auction and resale engine.
370 Agency
Full lifecycle delivery for agency clients with Next.js, Prisma, Express and PostgreSQL.
It started in a vocational school lab in Malang in 2016, with Android tutorials and a new GitHub account. Competition trophies followed, at school and then at Telkom University (20 placings, seven of them first places, from a chatbot prototype to data science and capture-the-flag). I helped lead a 200-student programming club and trained an IndoBERT model to read how students felt about their campus, while a part-time e-commerce backend paid for the late nights. Then came the production years: travel payments at MNC, multi-tenant SaaS at Jatis and digital-banking integrations at Jenius. In 2026 I joined Metrodata's Microsoft AI practice and changed how I build: I write the spec, send AI coding agents out in parallel and stand at the merge gate. Voice agents you can interrupt mid-sentence, a security platform that opens its own fix PRs, document AI that is not allowed to guess: every system ships with an evaluation that can say no.
Every feature starts as a written spec with acceptance criteria. AI coding agents (GitHub Copilot, Claude Code) do much of the typing in parallel, while I own the design, the review, verification on Azure and the merge. Evaluation is part of the product: gold sets, A/B tests, CI gates and shadow mode, and every number ships with how it was measured.
01
Every feature starts as a numbered spec with acceptance criteria and a constitution the agents must follow.
02
AI coding agents do much of the typing. I own the design, the review, verification on Azure and the merge.
03
Gold sets, A/B tests, CI gates and shadow mode. Negative results get written down and unhelpful changes stay off.
04
I build systems that can fail and recover, not ones that pretend to be perfect.
05
A metric ships with how it was measured (lab, offline, live) or it does not ship.
L1
L1
GH-600 · GitHub
Developing in Agentic AI Systems
Oct 2026(in progress)
AI-103 · Microsoft
Microsoft Certified: Azure AI Apps and Agents Developer Associate
Sep 2026
AI-901 · Microsoft
Microsoft Certified: Azure AI Fundamentals
May 2026
AZ-900 · Microsoft
Microsoft Certified: Azure Fundamentals
May 2026
IARC x APTIKOM
Microcredential: Associate Data Scientist
Dec 2021
Kominfo x IBM
Digital Talent Scholarship: Machine Learning (160 hours)
Jul 2021
HackerRank
Problem Solving (Basic)
Feb 2021