Hi, I'm Kiki.

Introduction

My full name is Mohammad Zulkifli Katili, but most people call me Kiki. I'm originally from Gorontalo, now based in Tangerang, in Greater Jakarta, Indonesia. I like problems where engineering, data and design meet, and I care as much about why something should be built as how.

Engineering journey

I studied Information Systems at Universitas Negeri Gorontalo, graduating as the best graduate of the Faculty of Engineering. My undergraduate thesis became Sindasi, a decision-support system using AHP-TOPSIS.

While studying I built production apps for real clients: Raisa, an inflation early-warning system for Bank Indonesia's Gorontalo office during my internship, then TimeMarket, a workload-management PWA for the same office. Project work followed — a nearest-mosque API and admin panel for Digimosque, a scheduling system for IAIN Sultan Amai Gorontalo, backend modules for Contag School, and spatial sales intelligence for PT. Makassar Raya Motor.

Current focus

I'm working on payment systems — payment gateways, B2B/B2C flows and the APIs around them — where correctness, idempotency and consistency matter more than anything clever.

As a Senior Fullstack Developer at Singapay I help maintain a payment gateway core that handles hundreds of thousands of transactions a day, on infrastructure that runs on Huawei Cloud with Alibaba Cloud as the disaster-recovery site. Day to day that means:

• Bank integrations under SNAP — Bank Indonesia's open-API payment standard, including the access-token and request-signature layer. Six banks directly, dozens more through vendors.
• One way to add a biller. I standardised the PPOB biller integrations — hundreds of service types and thousands of products — behind one internal schema and one abstract interface. Adding a new service went from about two weeks to about three days.
• Splitting the monolith. Busy transaction jobs moved onto asynchronous message queues, and new services are built in Go as parts of the PHP monolith are split out.
• Releases without a downtime window, using blue-green deployment.
• Tooling that keeps the ledger honest — nightly integrity checks, dry-run-first data patches that can always be reverted, and dispute-resolution tools that turned a 3–5 day case into a matter of hours.

I now handle the PPOB project, working from Indonesia with a team that is almost entirely in Shanghai.

Outside the day job I contribute to Anveesa, a suite of open-source developer tools (a database studio, database management, an AI retrieval tool and cloud storage), on both the front end (in Vue) and the back end. I'm involved in Anveesa Aras, Anveesa Vestra, Anveesa Nias and Anveesa Workflow.

AI, knowledge graphs & RAG

My master's in Computer Science at IPB University (Computational Intelligence & Optimization, GPA 3.83) focused on knowledge graphs and network analysis. The thesis became VektorPedia and a Scopus-indexed paper. Knowledge graphs are one of the most useful ways to give language models real context, which is why I keep exploring retrieval-augmented generation and LLM tooling. That interest turned into production work at Bank Indonesia, where I led two AI projects — see Leadership below.

Design background

Before I wrote much code, I designed. I led the multimedia division of the campus Linux Study Group, reached the finals of a national graphic design competition, and have designed logos and brand systems — HEXSTUDIO, Katili Corp, Kotaluwuk.com, and the GenBI Gorontalo identity. That habit shows up in the interfaces I build, including this site.

How I work

• Start from the problem. Understand who is affected and what "better" looks like before choosing a stack.
• Meet users where they are. Sometimes that's a PWA on the phone they already carry; sometimes it's an Excel import instead of an API.
• Be clear about ownership. On team projects, I say what the team built and what I built.
• Leave it maintainable. Conventions, small modules and readable code outlive clever ones.

Leadership

I have led the development of an AI application twice, both for Bank Indonesia's Regional Department in Jakarta:

• TIARA v3 (2025) — Bank Indonesia's AI policy chatbot. I led it end to end: architecture, features and deployment. Retrieval-augmented generation over PostgreSQL + pgvector with Gemini; a tool that lets the model query tabular data with pandas and return a summary; and a small self-hosted classification model (GLiClass, served on CPU) that screens every incoming prompt for injection attempts before it reaches the main model. Plus user and admin portals in React / Next.js and full technical documentation. The knowledge base reached 88% response accuracy, with 100% of feature tests passing, out-of-scope questions included.
• RAISA (2024–2025) — a risk-assessment system for transaction data reported by payment service providers (PJP) and money changers (KUPVA). Machine-learning prediction for the Net Risk Assessment (NRA APU) reaching 95% accuracy, an analytical dashboard over 50+ million transactions a month, and automation that cut manual analysis time by up to 30%.

Earlier, outside work, I chaired Generasi Baru Indonesia (GenBI) Gorontalo (2020–2021): 6 departments and 3 university commissariats, 60+ activities in a year, a rebuilt member scoring system that raised active membership to 80%, and the community's 2025 vision.

Publications

1. Katili, M. Z., Herdiyeni, Y., & Hardhienata, M. K. D. (2024). Leveraging Biotic Interaction Knowledge Graph and Network Analysis to Uncover Insect Vectors of Plant Virus. JISEBI. Link
2. Katili, M., Amali, L., & Tuloli, M. (2021). Implementasi Metode AHP-TOPSIS dalam Sistem Pendukung Rekomendasi Mahasiswa Berprestasi. Jambura Journal of Informatics, 3(1), 1–10. Link

Awards