Collaborated with the Global Address Manager (GAM) Team to enhance last-mile delivery accuracy in Brazil and Ireland. Supported initiatives to optimize address verification processes, leading to a measurable improvement in first-attempt delivery success rates across these regions.

Acted as the key point of contact to provide assistance for Proactive Station Launch Program (PSLP), focused on the Road Segment Model (RSM). Facilitated the correction of ~100K+ geo-spatial mapping errors ahead of delivery station launches, significantly mitigating downstream issues such as inaccurate routing and delayed deliveries.

Partnered in cross-functional PSLP engineering efforts to detect and correct geo-spatial errors using algorithms such as FB-AI (Facebook-AI-based Mapping) and GRIN (Generalized Road Network Inference), both of which integrated high-resolution satellite imagery to improve logistical path planning across North America.

Played a pivotal role in Amazon’s engagement with the Humanitarian OpenStreetMap Team (HOT-OSM), contributing geospatial enhancements for disaster-prone regions. This effort advanced the company's social responsibility goals while directly supporting disaster mitigation and risk preparedness through enriched mapping accuracy.

Collaborated on a three-member team that leveraged AWS S3 and internal mapping tools to store, process, and analyze geospatial data, enabling more efficient tracking and resolution of road network defects that previously hindered last-mile performance.

Pioneered a AI/ML initiative to improve Amazon's global product catalog quality by training image identification models capable of recognizing high-risk or misclassified products. The program achieved a False Negative (FN) rate of 19.59%, outperforming the 8% benchmark, and led to the successful classification of ~18.43 million ASINs that were previously undetected by text-based ML classifiers.

Developed robust, nested logic incorporating over eight distinct product attributes to accurately identify high-risk items from a database of 400M+ SKUs. This resulted in enhanced data traceability and strengthened compliance across diverse global marketplaces.

Achieved a classification accuracy rate of 99.5%, significantly exceeding the organizational benchmark of 98%, by leveraging sophisticated trend analysis and dynamic attribute-based risk stratification across the Amazon Catalog .

Provided critical technical feedback for machine learning pipelines deployed across eight international Amazon marketplaces, improving scalability and robustness of seller protection mechanisms .

Optimized daily operations by implementing streamlined SQL queries for these models and contributed to a 20% reduction in daily ETAs, directly enhancing fulfillment speed.