Developing advanced analytics approaches for fraud detection, collections enforcement, and audit selection. I create insights from customer data to improve compliance and workflow efficiency, deploy and monitor multiple mixed data-type models for fraud scoring with accuracy above 93% for UI claims, and retuned an existing fraud model to achieve 98% accuracy with an AUC of 0.99 for tax returns. I also formulated a unique semi-supervised audit selection model that achieved 94% accuracy, and my areas of practice include NLP (embeddings, LLMs, and RAG chatbots), recommendation engines, supervised data ETL (cleaning, feature engineering, and model tuning), and unsupervised approaches such as clustering and anomaly detection.

Working with students to master introductory college mathematics concepts. Preparing and lecturing college-level mathematics courses, including algebra and calculus.

Designing, developing, and improving mathematical models and practical implementations of target-tracking and computational-sensing algorithms (with the result that the required SNR for detection was halved), game-theoretic resource allocation and guidance planning (dynamic and adversarial assumptions), and spatial filtering of intermittent low-dimensional measurements for tracking accuracy and uncertainty quantification; also responsible for integrating these algorithms into operational and benchmark systems.

Developed novel algorithmic improvements to hashing, deduplication, load balancing, and VM migration systems. Achieved 300% efficiency gains in data reduction algorithms and optimized hash table performance in C for faster retrieval and reduced CPU contention. Applied machine learning techniques for predictive caching and streaming compression, while devising a convex relaxation to solve the NP-hard VM-to-resource assignment problem. Prototyped and tested all solutions using Python with numpy, scipy, scikit-learn, and related libraries.

Signal processing for software-defined radios. Developed methods for TDOA estimation and tracking of emmitters from multiple local receivers.

Developed and implemented algorithms for anomaly detection, predictive maintenance, and sensor fusion for industrial IoT applications. Led analytics research and development and defined new analytics products, including time-series anomaly detection, pattern identification, real-time data cleaning, device performance optimization, and infrastructure cost estimation tools. • Directed Python prototyping and algorithm integration within AWS/Docker, MongoDB, MQTT, and backend APIs. • Delivered predictive models with a 0.4% false-alarm rate (AUC of 0.92) on noisy, inconsistent data from 100k devices with 70 sensors. • Developed algorithmic improvements using numpy, scipy, and scikit-learn. • Granted patents related to the developed algorithms.

Research and development of radar signal processing algorithms for operational systems, with a focus on practical implementation and reliable detection performance. • In first month of employment, found critical bugs related to underlying statistical assumptions and behavior in the real-time signal-detector/characterizer and, for a particular input parameter setting, reduced the number of false-alarms by two orders of magnitude. • Designed and implemented (in C) a) solutions to detect and characterize radar pulse-compression codes (result: 50% reduction in missed-detections); b) Hough transform-based detection of overlapping chirp signals; c) robust statistical smoothing and outlier removal in C++, and supported integration into operational systems.

Lectured on signal processing and control theory; researched low-sample-support methods for correlation analysis of high-dimensional data sets.