At Amazon Web Services (AWS), I worked directly with enterprise customers to turn ambiguous requirements into production AI/ML solutions using SageMaker and Bedrock. I architected an intelligent document-processing pipeline that cut costs up to 40% and delivered 70–80% efficiency gains.
At AWS, I also resolved escalated SageMaker issues and monitored production models for drift and performance degradation. I ran distributed training jobs that cut execution time 50%, and tested endpoint variants to inform rollout decisions.
On a project, I built a private RAG pipeline for financial documents using Qdrant and vLLM, evaluating it with RAGAS and Hit Rate@K. Earlier, at Omdena, Inc., I worked on computer vision for wildfire detection and a DBSCAN-based credit-scoring system for underbanked Indonesian borrowers.
