Professional Experience

1. Accenture Indonesia

   Management & Technology Consulting — AI/ML, Data Science

   

   AI Decision Science Consultant (Industry & Functional AI)

   Jun 2024 – Present

   Full-time· Jakarta, Indonesia (hybrid)

   Shipping AI agents, voice AI, and deep learning systems across oil & gas, telecommunications, advertising, and FMCG engagements.

   • Data & AI Stream Lead, Satellite Methane Detection Platform for an oil & gas client (Jul 2026 – present): leading the stream for a platform going to production in October 2026 that fuses Sentinel-2, Landsat 8/9, and EMIT imagery with ground gas-sensor readings; PyTorch segmentation models and a Deep Agent that finds root cause and recommends next-best actions.
   • AI Engineer, Conversational Voice AI Platform for a telco client (2026 – present): building a self-hosted, real-time voice agent (STT → LLM → TTS) on Pipecat with a fine-tuned Indonesian TTS, Whisper on TensorRT-LLM, and Qwen3-32B on vLLM, deployed on AWS GPUs with Terraform.
   • AI Engineer, GenAI Ad Creation Platform, production SaaS with a Germany-based team (Sep 2025 – present): 35 merged PRs to the AI orchestration service, including the Campaign Brief, Template, and Creative agents (Pydantic AI deep agents with human confirmation before sensitive actions), RAG on Docling and pgvector, the LLM evaluation suite and CI gate, and image layer decomposition (LayerD, LaMa, OCR).
   • AI Governance Consultant, Responsible AI for an oil & gas client (Mar – Jun 2025): AI technology landscape assessment, technology selection guide, governance risk frameworks, and a Responsible AI testing toolkit.
   • Data Scientist, Retail Store Network Optimization for a telco client (Oct 2024 – Jan 2025): XGBoost demand model on H3 grid cells, integer programming and a genetic algorithm for store actions, K-means visitor segmentation, and LLM-generated store justifications.
   • AI Engineer, Customer Segmentation & GenAI Video Analysis for an FMCG client (Aug – Oct 2024): containerized RFM segmentation (77 K-means models) and data cleansing covering 5.8M customers, and a GenAI video-understanding solution that analyzes affiliate videos in minutes instead of hours.

   PythonAI AgentsLLMsRAGVoice AIPyTorchComputer VisionRemote SensingDatabricksAWSAI Governance
2. Delman Data Lab

   Data Analytics & Engineering

   

   Data Scientist

   Mar 2022 – Apr 2024

   Full-time· Indonesia (remote)

   Built data pipelines, predictive models, and reporting automation solutions for FMCG and manufacturing clients.

   • Reporting automation for an FMCG client — reduced operational costs by 50% and daily processing time by 99% (hours to minutes), with no human errors found after launch.
   • Procurement data standardization for a manufacturing client — unified 12K+ items, 1K+ suppliers, and 46K+ transactions into golden records and identified up to 50% potential cost savings.
   • Platform improvements, client demos, and delivery of forecasting, reconciliation, and predictive maintenance solutions.

   PythonSQLForecastingETLPredictive MaintenanceData Engineering
3. Daewoong Pharmaceutical

   Pharmaceutical & Healthcare AI

   

   AI Engineer

   Jul 2021 – Sep 2022

   Part-time· South Korea (remote)

   Developed AI solutions for pharmaceutical quality control and patient-facing health assistance.

   • Real-time encapsulation quality detection — EfficientNet on 16-frame image grids from 300K images, 93.5% test accuracy, shipped as a standalone desktop app beside the machine.
   • Sano, a Telegram health assistant — medication reminders (RASA NLU), drug identification from photos (MobileNetV2), and similar-drug recommendations (TF-IDF).
   • Korean customer-review analytics — BERT sentiment, BART summaries, and word clouds in a Flask + React dashboard.

   Computer VisionNLPChatbotTensorFlowSentiment Analysis
4. INSLAB, Pusan National University

   IoT Network Research & AI Systems

   

   AI-IoT Researcher

   Sep 2018 – Feb 2024

   Full-time· Busan, South Korea

   Conducted research on IoT network optimization, fault detection, and SDN-based resource allocation using AI/ML methods.

   • 6TiSCH network optimization using (multi-agent) Q-Learning — achieved +12% PDR, -23% energy consumption, -9% latency.
   • Wi-SUN IoT protocol implementation for smart utility networks.
   • Hydraulic system fault detection model — ~98% accuracy across multiple fault classes.
   • SDN-based task allocation framework for IoT-fog computing networks.

   Q-LearningIoT6TiSCHSDNPythonFault DetectionNetwork Simulation

Education

Pusan National University

Busan, South Korea

Integrated Master & PhD

Computer Engineering

2018 – 2024GPA: 4.28 / 4.5

Thesis

Improving 6TiSCH Network Formation and Transmission using Q-Learning

Key Coursework

Advanced Machine LearningIoT Network SystemsWireless Sensor NetworksAdvanced Computer Architecture

Institut Teknologi Sepuluh Nopember (ITS)

Surabaya, Indonesia

Bachelor of Computer Science

Computer Science

2014 – 2018GPA: 3.72 / 4

Thesis

Human Activity Recognition on Video Data using 3D Convolutional Neural Network

Key Coursework

Artificial IntelligenceComputer VisionData MiningSoftware Engineering