This is a living document. Use cases are documented progressively as part of an ongoing effort to make applied work accessible and reproducible. Check back for updates.
Flood segmentation from multispectral satellite imagery (ProCANet)
Satellite flood maps for 25M people, hours instead of days, state-of-the-art IoU.
Mining footprint detection with multi-modal satellite data
The first geospatial foundation model applied to continental-scale mining footprint detection.
Aquaculture pond detection and change analysis
Catching illegal shrimp farms from orbit before regulators arrive, hours, not field-survey years.
Flood policy evaluation, retention pond effectiveness in South Bandung
Do retention ponds actually reduce flooding? We answered it with satellites and causal stats.
Do Multilingual VLMs Abstain Under Cross-Modal Conflict in Low-Resource Languages?
A model that trusts its eyes in English becomes a caption-follower in Telugu. We measured it, localized it, and steered it back.
SEA-VL: multicultural vision-language benchmark for Southeast Asia
GPT-4V fails on Southeast Asian culture. We built the benchmark that proves it, ACL 2025.
GG-EZ: regional adaptation framework for vision-language models in SEA
Cultural AI for Southeast Asia without retraining from scratch, 98%+ global quality retained.
CommonLID: language identification on noisy web data
That 99% LID accuracy on SEA languages? It collapses on real noisy web data.
Share of Voice forecasting system, Fortune 500 APAC (Artefact)
Turned media gut-feel into calibrated prediction intervals across 6 APAC markets for a Fortune 500.
Biometric authentication and alternative credit scoring at scale (GDP Labs)
1M+ daily inferences at 99.99% uptime. Credit for the borrowers banks wouldn't score.
Real-time fraud detection pipeline (GDP Labs)
Three fraud attack types, one pipeline, rule filters, XGBoost, and graph anomaly detection.
Municipal waste logistics forecasting, Jakarta Smart City
Forecasted Jakarta's trash so trucks follow where waste actually piles up instead of a fixed weekly schedule.
Demand and audience forecasting for financial and media clients (Artefact)
Two Fortune 500 clients, two forecasting problems, two weeks each, shipped and handed off.
HakkTaxi: ride-share demand prediction (Microsoft Azure APAC Hackathon)
A Jakarta ride-demand heatmap built in 48 hours. Microsoft Azure APAC Regional Champion.
TeleHealthMonitor: Edge AI for remote patient monitoring (CamvsCovid)
COVID-19 vitals monitoring on a 2G connection, no cloud, no smartphone. Top 3, Cambridge.
Community IVR: Voice AI for offline communities (Cal Hacks 8.0)
Health info for 40% of Indonesians without smartphones, via a plain old phone call.
Plastic bag ban, causal policy analysis using NLP and citizen data
Jakarta banned plastic bags in 2020. Did it work? Proved it causally with 100K+ complaints.
Qwen VL Fine-tuning for AI City Challenge 2026 Track 2
Fine-tuning a 3B VLM on a 14GB GPU, every memory constraint hit, diagnosed, and solved.
Conformal Prediction for Vision-and-Language Navigation (DUET + HAMT)
21 uncertainty methods for robot navigation tested. Most fail. One works, and we know why.
Contributing to llm-d/inference-scheduler, Kubernetes LLM Inference Scheduling
Contributing to the Kubernetes layer routing LLM requests across GPU backends at production scale.