May 2026 – Present
Head of AI Engineering
BukuWarung
I build the internal platforms that let teams across the company automate their workflows. We are currently building three of them: an AI agent factory, an AI voice studio and an agentic CRM.
May 2026 – Present
BukuWarung
I build the internal platforms that let teams across the company automate their workflows. We are currently building three of them: an AI agent factory, an AI voice studio and an agentic CRM.
Oct 2021 – Apr 2026
Bank Rakyat Indonesia · Jakarta, Indonesia
I headed BRIBRAIN, the department that builds BRI’s AI products. The team works on customer engagement, fraud and risk analytics, credit underwriting, and automation for branch and contact-centre operations.
Highlights
Sabrina is BRI's AI assistant. It answers customers on WhatsApp and in the bank's app. According to BRI's 2024 annual report, it handled 39.6% of the messages sent to the Contact Center, and customer satisfaction rose from 69.8% to 80% over the year.
Anugerah Inovasi Indonesia 2024, held at the Indonesia Stock Exchange in Jakarta, recognised BRI's AI assistant Sabrina for innovative products and business models in virtual services and anti-fraud.
At the Ministry of State-Owned Enterprises' AI Center of Excellence kick-off, BRIBRAIN showed Sabrina's conversational banking, the WISE generative-AI search engine, and robotic cash management to 15 state-owned enterprises.
Tribunnews' Mata Lokal Award recognised Sabrina's use of large language models to answer customers in Indonesian regional languages.
Bank Service Excellence Monitor is an annual survey of Indonesian banks by the research firm MRI and Infobank magazine. Sabrina came first for chatbot performance, with a score 16.28 points higher than the year before. BRI's mobile app, BRImo, came first for mobile banking in the same survey.
Under the agreement, Sabrina moved to Azure OpenAI Service so it could understand Javanese, Sundanese, Padangnese and other regional languages. WISE uses Azure Cognitive Search.
WISE (Work Instruction Search Engine) is a generative AI tool built by BRIBRAIN. Branch and contact-centre staff use it to find the right procedure when they handle a customer's question. It won BRI a second IDC Future Enterprise Award in a row.
BRIBRAIN, BRI's AI and machine learning platform, won Best in Future of Intelligence at the IDC Future Enterprise Awards 2023, for both Indonesia and Asia Pacific. Its recommendations help the bank's field staff decide which customers to approach and what to offer them.
Dec 2020 – Oct 2021
Nomura Research Institute
I worked on computer vision projects. One of them was a camera-based system that approves refuelling at self-service petrol stations, which Tatsuno later released in Japan as AIGO.
Aug 2020 – Dec 2020
gaw.ai
I started gaw.ai to apply computer vision to the engineering and inspection work I had done before my PhD. We built software that reads engineering drawings, and a tool that generates corrosion loops automatically.
Highlights
A web app that groups piping components into corrosion loops, using k-means clustering and the group technology method. It takes an Excel sheet of operating conditions and materials, and returns the same sheet with a loop assigned to each component. It implements my 2020 paper in the Journal of Quality in Maintenance Engineering.
The template matching service behind gaw.ai auto-label. It is written in Python and served with Ray across a head node and worker nodes. It was written by an engineer on the gaw.ai team.
The software detects and recognises the symbols and text in piping and instrumentation diagrams (P&IDs) and other engineering drawings. It then exports equipment, instrument, valve and line lists to Excel or CSV, and can digitise legacy drawings.
The products were built for engineering teams in oil and gas and petrochemicals. Auto-label detects and recognises symbols in engineering drawings, auto-text searches the information in them, and CL generates corrosion loops.
Jul 2019 – Jul 2020
GetGo
I co-founded GetGo with Erdian Tomy, who was CEO, and ran the AI side. We built cashierless kiosks that recognised items with computer vision, and later a visual search engine for fashion e-commerce.
Highlights
COVID-19 closed the offices where GetGo's cashierless kiosks were installed. We reused the same image recognition technology to build a visual search engine, which matches a photo of an outfit to products in a retailer's catalogue.
An article on GetGo's two products: a cashier-less mini store for offices, and a visual search API that lets shoppers find fashion items by taking a photo inside a retailer's app.
GetGo was one of eleven startups in the fourth batch of Gojek's accelerator. The programme had trained 35 companies since it started in 2019.
May 2016 – Apr 2019
University of Stavanger · Stavanger, Norway
My research was on machine learning and knowledge representation for risk-based inspection in oil and gas. My supervisor was Prof. R.M. Chandima Ratnayake. The thesis is titled Artificial Intelligence Approaches to Lean Risk-Based Inspection Assessment.
Highlights
An EU project with universities in Italy, Finland and Norway. It looked at Building Information Modelling as a training subject, with the aim of improving training tools and professional qualifications and supporting apprenticeships.
Awarded at the 28th International Ocean and Polar Engineering Conference in Sapporo (the ExxonMobil Prize), for a paper on neural network models for risk-based inspection screening. I wrote it with Prof. R.M. Chandima Ratnayake.
An EU project with universities in Italy, Poland, Norway, Portugal and Finland. It developed new ways to teach lean thinking for office and knowledge work.
Jul 2011 – Jul 2015
Rekayasa Solverindo · Wood · TechnipFMC · IKM Operations · Indonesia and Norway
I worked in asset integrity for four years, first in Indonesia and then in Norway. Most of it was risk-based inspection for refineries and for onshore and offshore production facilities. I also did material selection for subsea equipment, and inspection engineering for an offshore platform.
Highlights
Maintenance and inspection engineering for Total's Martin Linge platform, with IKM Operations. We coordinated with Samsung Heavy Industries, which built the topside, and Technip, the main contractor. The work also included operational support and general maintenance.
A summer project at FMC Technologies (now TechnipFMC). We developed a concept for a subsea heat exchanger that runs at high temperature in a corrosive environment. We looked at two cases: cooling the well stream with seawater, and exchanging heat between the well stream and a separate medium.
I was part of the Wood team that implemented risk-based inspection for all Santos facilities in Eastern Australia, from 2012 to 2013. The work covered corrosion loops, inventory groups, integrity management plans and integrity operating windows. We delivered a risk level for each facility and an optimised inspection plan, following API 581 and using the APIRBI software.
I was part of the Wood team that assessed four offshore platforms and three onshore processing plants for HESS. We developed corrosion loops and inventory groups, and delivered risk levels and an inspection plan based on API 581.
A risk-based inspection assessment of the Fuel Oil Complex and the Lube Oil Complex at Pertamina's Refinery Unit IV, using DNV's Orbit Onshore software. I did this work at Rekayasa Solverindo.