Open to work

7 years running ecommerce operations. Now I build the tools — and audit them the way an operator would.

Achmad Khairul Anam — Amazon, Etsy, Shopee, 15+ seller accounts, P&L owned end to end. Today I ship AI products solo, and I check my own work the way I'd want a vendor to check something I'm paying for.

See the audit log → Get in touch
Audit log — 001 / 004
2026 · Marketplace lead-gen infrastructure
Revenic.io
Live →

The problem: AI-generated sales data is only useful if the numbers can be trusted. One hallucinated lead score can cost more in wasted sales effort than the tool saves.

Trade-off Chose a "Deterministic First, LLM Last" architecture — every score, match, and flag stays rule-based and auditable. The LLM is restricted to narrating results in plain language, never computing them.
Status Pipeline built (Apify scraping + Hunter.io + MillionVerifier for verified contacts). Currently in live testing — validating end-to-end lead accuracy before wider use.
Next.jsTypeScriptApifyClaude API
2026 · CV-to-job matching, live
DiterimaKerja.id
Live →

The problem: a job-matching tool was live and converting paying customers — but AI parsing reliability was silently degrading behind a working interface.

Found Three separate root causes: a silently-deprecated model breaking every parse, a 4,000-character input limit quietly dropping work history from longer CVs, and unreliable JSON from a reasoning model — none visible without deliberate investigation.
Fixed Switched to a non-reasoning model for structured extraction, added defensive JSON repair. 18+ paying customers since launch, zero ad spend, fully organic.
Node.jsOpenRouterClaude Sonnet 5REST APIs
2026 · Amazon/Etsy brand protection
Defenr.com
Live →

The problem: a platform that detects hijackers for other sellers only works if its own infrastructure is trustworthy.

Found A public authentication exposure in my own infrastructure, during a routine audit — no external report, no incident.
Fixed Closed it with HTTP Basic Auth middleware, owned the full audit-to-fix cycle solo, verified before calling it done.
Next.jsTypeScriptSupabaseVercel
2026 · Shopee product research
CekLaris.com
Live →

The problem: before wider rollout, I audited my own live product instead of waiting for a user to find something first.

Found A premium-payment flag trusted entirely on the client side — exploitable from the browser console — plus a usage counter that overstated real traction.
Fixed & verified Rebuilt payment verification server-side, re-ran the original exploit myself to confirm it's closed. Also relabeled every AI-estimated result honestly and added a 7-day money-back guarantee.
OpenAI APIVercelMidtrans
About

7+ years running multi-platform ecommerce operations — Amazon, Etsy, eBay, Shopee — owning P&L across 15+ seller accounts, including my own Amazon FBA and an Etsy shop I still run today.

Now I build AI-powered products solo, using the same instinct: find what's actually broken, fix it, verify it works. I use AI tools to move fast, but I test everything before I trust it — every story above is a bug I found in my own shipped work. That's not an accident. It's the habit.

Contact

Open to Founding Engineer / Product Engineer roles at early-stage startups, ecommerce-tech & martech companies, Technical PM / Solutions Engineer roles, and freelance data & automation projects.

achmad.khairul@gmail.com WhatsApp LinkedIn GitHub Apify