Ardacandra Subiantoro

Architected and deployed a suite of credit models (Income Prediction, B-Score, Collection Prioritization) on Google Cloud Platform using XGBoost and LightGBM to optimize credit assignment and recovery strategies for millions of users. Built a real-time risk system with sub-second inference speeds, improving Risk-Adjusted Returns by 30%. Led ML infrastructure migration from AWS to GCP and developed scalable pipelines in Python and BigQuery, accelerating training cycles from 3–5 days to 1 day. Implemented automated model observability dashboards (PSI, prediction distributions, ROI/CLV) and mentored 2 junior data scientists.

Designed dynamic pricing strategies based on historical booking data, driving occupancy revenue. Executed data quality and ETL workflows to ensure high-integrity datasets for downstream reporting and decision-making.

Computer Vision (Surveillance): Engineered a CCTV face tracking pipeline using RetinaFace and detection-based tracking; improved face detection recall by 4.47% in low-illumination environments via a custom post-processing algorithm. NLP (Biomedical Indexing): Implemented Latent Dirichlet Allocation (LDA) for unsupervised topic modeling to automate MeSH indexing, achieving 74% topic similarity with human labels.