Senior AI/ML Engineer

I joined eBay's Marketplace Intelligence team to deliver production AI solutions for search discovery, seller support, and personalization across European marketplace operations. I worked with machine learning engineers, product managers, and marketplace developers across six stakeholder groups to design recommendation models that improved related-item click-through rate by 14% across three high-traffic buyer discovery surfaces.

I built RAG assistants using LangChain, embeddings, and vector search over internal documentation for seller-support workflows, validating response quality across more than 1,200 reviewed queries. I also fine-tuned transformer-based classifiers for listing categorization, improving offline F1 by nine points before controlled experiments.

I deployed FastAPI inference services with ONNX optimization behind Kubernetes workloads handling more than 40,000 predictions per minute. I established MLflow and SageMaker workflows for training, model registry management, and controlled deployment, shortened release cycles by 28%, implemented drift and latency monitoring, and mentored engineers on RAG evaluation, prompt testing, and responsible generative AI practices.

I developed predictive analytics solutions for shipment delays, ETA forecasting, and operational decision support across international cargo networks. Working with logistics analysts, software developers, and operations planners, I built delay-risk and ETA prediction models using Python and Scikit-learn that helped planners prioritize over 500 monthly consignments.

I built Spark and Airflow pipelines that combined port, carrier, and shipment attributes from 12 source tables, reducing manual preparation effort by 60% for recurring model training. I deployed batch inference workflows to identify high-risk shipments and enable earlier customer communication during disruptions.

I also experimented with NLP classification for free-text EDI and booking exception notes, reducing manual incident categorization time by 41%. I tracked experiments in MLflow and documented model versions across eight iterations to strengthen governance and review processes.

I created SQL datasets and Python analytics services for manufacturing intelligence products, providing visibility across 20 production lines and supporting plant-level operational decisions. I collaborated with product managers, consultants, and software teams on ERP data related to production planning, machine utilization, and workflow management.

I developed Scikit-learn prototypes using historical job and work-center data that improved duration estimates by 16% and helped customers identify scheduling risks earlier. I also created FastAPI services that delivered KPI information to React dashboards used for more than 200 daily views by manufacturing supervisors and business users.

I automated nightly data preparation jobs to reduce daytime database contention by 24%, while creating reusable datasets for reporting, analytics, and future machine learning experiments.

I supported the expansion of a digital wallet platform, contributing to transaction services, QR payments, remittance workflows, and merchant-facing financial applications. Working under senior engineers, I built Python and SQL utilities around wallet transaction history and payment reporting for services used by hundreds of thousands of Filipino consumers.

I prepared reporting queries and data scripts that helped risk and product teams analyze payment patterns and failed remittance cases, reducing manual investigation effort by 33% across two business groups. I also supported production debugging, API maintenance, and event-data preparation for customer-facing payment services.

My work contributed clean datasets for four product campaign reviews, helping teams evaluate feature adoption and customer behavior.