- Led the design and implementation of an internal Information Retrieval (IR) system leveraging AWS OpenSearch, exploring and optimizing various search methodologies, including lexical, dense vector, hybrid, and sparse neural search, to maximize system performance.- Spearheaded the application of RAG-based knowledge extraction techniques across multiple enterprise use cases, resulting in a substantial improvement in user satisfaction as measured by post-interaction surveys.- Developed and deployed an NLP-based solution for validating metadata in internal data assets, ensuring alignment with enterprise compliance standards while providing actionable feedback to users for data quality improvement.- Collaborated with cross-functional teams, including Product and Design, to define strategic objectives and drive the creation of high-impact machine learning solutions, enhancing user experiences across a broad range of models.

- Designed and implemented a neural network-based recommender system for enterprise users, working closely with the software engineering team to successfully deploy the model into production, ensuring scalability and performance.- Developed and trained a custom two-tower reranker model to personalize search results for enterprise platforms, integrating both graph-based and text embedding models to improve NDCG metric.

- Developed an HMM based predictive model utilizing customer transaction and credit bureau data to identify key marketing variables to improve campaign effectiveness.

- Developed large-scale models for production cost analysis of power grids, focusing on the integration of renewable energy resources to support grid optimization and planning studies.- Conducted data validation through advanced statistical techniques, ensuring the accuracy and reliability of measurements collected from sensors deployed across the distribution grid.- Developed probabilistic models to generate synthetic load data, enhancing the validation of power grid models by supplementing actual measured data for more robust simulations.- Collaborated on multi-disciplinary projects focused on the development of microgrids and the control of distributed energy resources (DERs), contributing to innovative solutions for grid stability and efficiency.

Taught EE 342: Probability and Statistics class. Developed course syllabus, class lectures, and taught a class size of 53 students.

• Machine learning methods for real-time anomaly detection: Developed an online Kernel Density based outlier detection algorithm for bad data detection in the power grid sensor data in a decentralized manner. To further improve computational efficiency, proposed an online sparse least-squares one-class support vector machine algorithm for the real time bad data in the measurement data and also demonstrated the efficacy of the algorithm in the detection of false data injection attacks. • Sensor placement problem in the power grid: Developed algorithms for placement of sensors in the power grid network and obtained performance bounds for the proposed algorithms using the generalized eigenvalues of matrix pencils.

Taught EE213 (Basic Circuit Analysis II) labs in Fall 2011 and Spring 2012 semesters.

Dimension reduction in approximate Bayesian computation: Worked in the Mathematical Informatics Lab on the application of sufficient dimension reduction techniques to the approximate Bayesian computation (ABC) to increase computational efficiency.

Worked on identification of fault location in cascading failures in the power grid.

Worked on bad data detection in power grid.

Carried out periodic checks of core network elements to meet required KPI values and ensured proper management and dynamic dimensioning of voice and data traffic in own network, as well as at the points of interconnection.

Monitored the radio parameters, analyzed drive test measurements, made recommendations for adjustment of various network parameters, as well as necessary capacity expansion to improve network quality.