Resume

Own ranking and recommendation improvements for Facebook Feed, serving billions of daily users

Independently shipped 10+ production ranking launches across comments, unique commenters, profile visits, and creator-friend interactions

Led 79+ ranking quality experiments using NDCG-driven evaluation, offline validation, and large-scale A/B testing

Increased coverage of a key engagement training signal by more than 200% through an end-to-end logging pipeline audit

Led creator distribution fairness work using debiasing, label reweighting, and propensity strategies that increased follow events by ~4%

Migrated production prediction tasks from single-task models to multi-task learning architectures with task-specific adapters

Led applied AI research in computer vision, multimodal learning, LLMs, and human performance assessment

Built multimodal systems fusing video, eye-tracking, and language to evaluate trainee performance with real-time feedback

Integrated LLMs into simulation workflows to generate context-aware coaching and assessment summaries

Designed a virtual simulation environment for rapid prototyping of training and assessment applications

Built infrastructure for ML data ingestion and model-support workflows

Designed real-time data ingestion and delivery systems for modeling, analytics, and operational ML workloads

Developed cloud-native services and CI/CD pipelines on AWS to improve deployment reliability and developer velocity

Developed ML algorithms for automated performance evaluation from multimodal imaging, gaze, and video data

Built data analysis pipelines for multimodal research datasets, including preprocessing and public release preparation

Conducted research in video understanding, human activity analysis, and explainable AI

Researched predictive modeling, pattern recognition, anomaly detection, and text mining

Explored active learning and scalable analytics for extracting patterns from large, complex datasets

Led a training management platform serving ~600 users across 12 countries

Built systems to evaluate staff skills by role, identify capability gaps, and recommend training plans

Implemented scheduling and resource-management modules for training programs