Oct 2023 – Present · 2 yrs 7 mos · Ahmedabad, Gujarat, India · On-site

• Created an evaluation framework using WeatherbenchX to benchmark Client's ML weather models (Graphcast, Gencast, etc.) against public (Mesonet, ISD, ECMWF) and private datasets, ensuring accuracy across multiple atmospheric variables.
• Built geospatial data pipelines that transformed Earth Engine and Zarr data into ML-ready datasets with advanced coordinate transformations and dynamic resampling.
• Architected a thread-safe caching system that reduced Google Earth Engine API calls by 40% and cut query latency by 90% (5s → 500ms).
• Developed Xee, a high-performance data retrieval tool leveraging adaptive request partitioning and parallel processing for large-scale datasets.
• Core maintainer of https://github.com/google/weather-tools, https://github.com/google-research/arco-era5

Jan 2023 – Sep 2023 · 8 mos · Ahmedabad, Gujarat, India · On-site

• Redesigned the ARCO-ERA5 pipeline, creating a unified framework to process petabyte-scale weather datasets, achieving 85% faster data ingestion into BigQuery & GCP Buckets (recognized with client appreciation).
• Conducted a POC on India’s commodity transport flows (wheat, rice, and others) to evaluate barge transport as an alternative to trucks/trains, uncovering opportunities to reduce logistics costs and improve efficiency.