About
I build vision systems that run on real photographs
I came into machine learning sideways: a computer science degree at Telkom University, a computer-vision bootcamp at Indonesia AI where our team placed second, then jobs where the models had to hold up. Junior AI engineer first, production vision systems after that, and now AI engineer at ENB Mobile Care, where a six-model ensemble grades used smartphones and the grade decides what someone gets paid for their trade-in.
That last part shaped how I think about the work. When a model's output moves real money, top-1 accuracy on a clean test set stops being the number anyone cares about. What matters is what the thing does with a blurry photo, when it should refuse, and who finds out first if it degrades quietly. Most of what I build goes at those questions: quality gates, calibrated abstention, evaluation reports someone else can reproduce, health endpoints and monitoring.
Outside the day job I run my own systems, mostly so the habits stay sharp. LeafSentry serves a public inference API and abstains when the photo is not good enough to triage. Axiom Alpha is an equity-analysis platform I built and operate end to end. KenalinIkan is an Indonesian fish encyclopedia covering 130 species with regional names, season maps, and market price ranges. SIMBK is counseling-records software a school actually uses, where an audit journal and a backup restore that has been rehearsed count for more than model architecture.
I am open to AI engineering, machine learning, and computer vision roles. The demo page runs a real model if you would rather look than read, and the evaluation page has the report behind it.