Masa sih, Jar?
I direct AI to build production systems, then use my QA brain to find what could fail before users do.
A bit about me
Muhammad Azhar Maulana
B.Eng, Computer Engineering, ITS · GPA 3.42 · Distinction
Masasijar comes from a question I keep hearing: “Masa sih, Jar?” This site is my answer.
I'm a junior developer with a QA foundation. Testing taught me how systems break, which is unusually useful when I direct AI to write code.
I don't hand-write every layer. I specify the behavior, use AI to build, then attack the result like it came from someone else. Flutter apps, Next.js sites, agent workflows, nginx configs: they still have to survive production.
I'm not claiming deep expertise in every tool listed here. I'm claiming I can take an idea to production, verify it, and stay accountable for the result.
Selected work
Production systems, internal tools,
and the testing that holds them up.
A WhatsApp agent that handles the first layer of product questions for a laboratory instrument distributor. It verifies context, answers from the available product information, and hands the conversation to a human when a client asks for a quotation.
Guardrails enforce identity checks and mandatory human escalation before quotation follow-up or other outbound replies.
A multi-platform internal application for quotation and invoice creation, leads management, inventory, and the AI Customer Service dashboard. Role-based access keeps web, desktop, and mobile workflows aligned.
The Flutter client uses self-hosted distribution and version checks, with CI/CD plus automated exchange-rate ingestion through Supabase Edge Functions and pg_cron.
A multi-tenant platform where hosts collect guest-perspective event photos. Guests join through a QR code and receive an upload slot without creating an account.
In development with media storage and delivery across Supabase and Cloudflare R2.
A mobile-first learning app for Indonesians aged 40+: five guided levels, an AI image detector, a forward checker and a prompt translator. Directed and shipped with AI under a hard deadline.
A small system for recurring bills, subscriptions and Spotify Family cost splitting, running behind nginx on my own VPS.
End-to-end testing across tracking, reporting, and dashboard releases. I found substring matching that inflated brand mentions, flagged context loss in regex sentiment classification, and profiled API degradation under concurrent load.
QA field test
A deliberately buggy demo. No real orders or payments.
Test item: 10 credits each
Try the checkout and watch the receipt log.
Trigger a suspicious action to reveal this explanation.
Trigger a suspicious action to reveal this explanation.
Trigger a suspicious action to reveal this explanation.
Capabilities
Clear about what I know.
Honest about where AI helps.
The part I actually trained for.
Agent design, not just prompting.
Specified by me. Built with AI. Tested like an adversary.
These layers support day-to-day business operations. They have to hold up after the demo ends.
Where the soldering iron still wins.
Along the way
Avonetiq, Jakarta
Sribu.com, Jakarta
Digital Skola
Bangkit Academy, GCP
From the journal
What I learned while making, testing, and shipping.
Off the clock
Music is part of my life, too.
Here’s what I’ve been sharing as ajarrr.
Loading the latest tracks from SoundCloud…
A playable detour
I also play games. Pick one and stay a little longer.
Move, survive for 60 seconds, and let one shield absorb one collision.
Arrow keys or A/D. Hold the on-screen controls on touch.Switching games starts a fresh run.
Let’s talk
QA, AI automation, or a product that needs to survive production. I’m early in my career, honest about my process, and ready to do the work.
Open to work & contracts
Score is 10 points per second plus 25 per bug passed. Best: No record yet
masasijar
