New York City Metropolitan Area
Staff Machine Learning Engineer | Recommendation Systems Architect | PyTorch | Personalization & Video Recommendations I am a Machine Learning Engineer with a 10+ year track record of architecting and deploying large-scale, business-critical machine learning systems. My expertise lies in the design and optimization of Recommendation Engines for Video and Live Event platforms, where relevance, ranking, and user engagement are paramount. Throughout my career, I have bridged the gap between cutting-edge ML research and production-grade software, delivering solutions that drive tangible business growth. I specialize in solving the unique challenges of recommender systems, including cold-start problems, temporal dynamics, and multi-objective ranking. I conducted numerous A/B tests across different regions, improved total watch time metrics. Core Competencies: Recommendation Systems: Expert in designing end-to-end discovery platforms, utilizing Transformer, GNNs and Deep Learning (DLRM) to personalize content for millions of users. Video & Live Events: Specialized experience in modeling user interactions in live-streaming environments, optimizing for watch time and retention. Technical Leadership: Proven ability to lead engineering squads, mentor senior talent, and define technical roadmaps. System Design & Scale: Architecting high-throughput ML pipelines (Python, Spark, Kafka) that handle massive scale. Technical Stack: Machine Learning: PyTorch, TensorFlow, Scikit-learn, Keras, XGBoost. Algorithms: Transformers, GNNs, Learning to Rank, Vector Search for user and title embeddings. Infrastructure: Python, SQL, Docker, AWS/GCP, Faiss, Vector Databases. Data Engineering: Apache Spark, Snowflake, Kafka, Airflow. I am passionate about building the next generation of personalized discovery engines. If you are solving complex problems in the RecSys space, let's connect.
• Architected personalized recommendation system for 100+ million users using transformer (SASREC) and session based graph neural network models (SRGNN). • Architected live sports ranking and retrieval models using transformer based sequential model. • Conducted A/B tests across different regions, improved total watch time metric. • Built a RAG for live events popularity prediction using Gemini Pro model, Google search grounding and vector search (ANN) for event embeddings. • Scaled a RAG for story plot of titles in Disney+ catalog and augmentation for genre and maturity rating using Gemini Pro model and Google search. Spearheaded content understanding and content relevance using content embeddings. • Scaled data pipelines using Databricks Spark for User Modeling & Profiling
• Deep learning: Model offline evaluation and A/B testing using PyTorch and MLOps tools. • Feature Store/Data Pipeline Ran feature importance jobs, updated feature store and automated data pipelines. • Model Iterations: Setup opt-out models to mitigate the impact of privacy updates launched by Apple. • Cross Functions: Helped product managers with the roadmap by predicting the critical path and roadblocks.
• Machine Learning: Bond similarity service • Rest API: Used AWS Lambda to bring cached scenario results. • Quant: Marked CLO portfolios to market, get daily return series. • End to End Systems: Built a large-scale system to support 20 investment professionals. • Data Engineering: Built database using Intex and IHS Markit APIs for leveraged loans.
• Management: Led a team of 4 quantitative analyst • End to End System: Large-scale analytics system for 20+ portfolio managers. • Machine Learning: Estimated a risk-adjusted student loan rate model using logistic regression. • Software Engineering: Code reviews, automated tests and production change review. • Data Management: Built a Data Lake and Data Warehouse with residential/commercial mortgage data. • Data Pipelines: Bloomberg, Intex, CapIQ, Markit Skills: Machine Learning, AWS, scikit-learn, Python, Pandas Sql · Distributed Computing · Model Training ·
• End to End Systems: Trading tools for Non-agency and Agency RMBS trading desks. • Data Analytics: Inferred loan modifications using 1010data • Model Backtesting: Compare model projections and actual performance using C++. • Model Development: Non-agency model implemented in C++.