Rashmi Singh

I’m a Computer Vision Engineer specializing in large-scale video/image analytics, domain adaptation, and edge-deployment-aware pipeline engineering. I built an end-to-end three-stage ML pipeline processing 1M+ street-level images using sequential transfer learning (PA-100K → PETA → Jakarta) and model ensembling (EfficientNet-B3 + YOLOv11M), including manual ONNX inference pipelines for production deployment without framework wrappers.

I focus on data quality and field validation: I curated a Jakarta-specific dataset with 8,400+ annotations in CVAT (κ=0.87), achieved 73% correlation with ground-truth via systematic manual and 48-site field validation (r=0.73), and added transparency with a Spatial Uncertainty Index. Alongside this, I engineered large-scale image collection and aggregation using Google Street View API, and earlier I supported HVAC pre-sales work by automating risk assessment and forecasting workflows, reducing manual review time by 40%.