Greater Seattle Area
→ Machine learning engineer and applied scientist with experience in building end-to-end production-grade models and scalable ML systems. → Specialized in machine learning algorithms, deep learning, optimization, and applied statistics. → Author of the Springer book Hands-on Deep Learning: Building Models from Scratch; numerous peer-reviewed publications and patents. → Stack: Python, PyTorch/TensorFlow, NumPy/Pandas, Scikit-Learn, Spark, Airflow, AWS, Docker.
Leading machine‑learning efforts for Okta’s AI-based product offerings (Okta AI): • Identity Threat Protection (ITP) Risk Engine — continuously scores user and session risk to stop identity‑based attacks in real time. • Bot Protection — detects credential‑stuffing and automated abuse before it reaches customer applications. • Toll Fraud Mitigation — blocks telephony‑based IRSF (international revenue‑sharing fraud) and other voice‑authentication scams. • Identity Governance Analyzer — provides fine‑grained access‑review analytics to ensure the right users have the right permissions.
• Full data science life-cycle from conception to building production-grade machine learning systems through model prototyping, training, tuning, deployment, and monitoring. • ThreatInsight threat scoring algorithm with a gradient boosting classifier and deployment through PMML artifact conversion. • ML-based “reputation-scoring” service using Neural Nets, XGBoost, Embeddings, AWS SageMaker, Glue/Spark, Snowflake, Step Functions, and Docker. • Toll fraud detection and mitigation technique through unsupervised isolation forest algorithm. • Patent pending inventions in the areas of Machine Learning, Generative AI, and Security.
• Advanced science, optimization, and machine learning algorithms for satellite orbiters and airborne instruments. • Intelligent deep learning-based systems (CNN, RNN, LSTM) for MARS autonomous rover. • NASA’s data intensive and data-driven science systems for satellite missions. • Automated Machine Learning (AutoML) workflow for microwave instrument data science. • Unsupervised anomaly detection and signal processing on time series streaming sensor signals.
• Next-generation optimization algorithm for generating NASA’s thermal infrared data product. • Python evaluation and data visualization tools to monitor performance metrics at scale.
• Multi-layered neural networks development on satellite datasets and integration into NOAA’s operational FORTRAN-based system. • Bayesian variational optimization algorithm in an integrated satellite retrieval system. • Algorithm deployment on NOAA’s 4,700-core supercomputing facility for parallel processing. • Data assimilation and numerical weather prediction (NWP) modeling.