Pravi D. Email & Phone Number

Charlotte, North Carolina, Us

- Leading the development of innovative AI solutions in cybersecurity and nuclear sectors, leveraging deep domain knowledge to drive growth and address complex challenges.- Proficient in developing and deploying advanced NLP models for text classification, entity recognition, and topic modeling using cutting-edge techniques like BERT and its variants.- Demonstrated strong MLOps skills by managing the entire lifecycle of AI projects, from implementing automated retraining pipelines and real-time reporting dashboards to ensuring alignment with business goals and regulatory compliance, ultimately enhancing model performance and operational efficiency.- Experienced in working closely with product managers, engineers, and data scientists to integrate AI solutions into existing platforms, fostering innovation and efficiency.- Actively mentoring emerging data scientists, promoting a culture of learning and professional growth within the team.

Charlotte, North Carolina, Us

Work with business units across Duke Energy to interact with stakeholders, translate business problems into data problems and address them using machine learning and AI. • Develop NLP methods to summarize and search large-scale text data.• Design knowledge graphs to extract explicit relationships across organizational policies.• Develop risk model and optimized vegetation trimming schedules for electric distribution to reduce vegetation-related outages using.

Oak Ridge, Tn, Us

Led and collaborated on several projects funded by Department Of Energy (DOE) and Department of Science (DOS) offices. Working primarily on machine learning applications for urban sciences and materials science.• UʀʙᴀɴNᴇᴛ: Used network science to quantify complex interdependencies in critical infrastructure. Tool is not integrated into FEMA's situational awareness tool EAGLE-I. • Used NLP to summarize and cluster research publications to recommend potential candidates for open research positions.• Chaired the SMC Data Challenge 2021 - 8 research challenges and 23 teams from across the world.

Oak Ridge, Tn, Us

• Compressed convolutional autoencoders using tensor decomposition methods to 20% of its original size with minimal loss of accuracy for microscopy images.• Designed DNNs for object detection for autonomous vehicles and reduced inference latency inYOLOv2 by 22%.• Developed SVM regression models that fuse sensor data from nuclear power plants and used random forest classifier to identify sensor faults.

Riverside, Ca, Us

My PhD dissertation focused on developing algorithms to identify spatial and temporal changes among social media and smartphone users. This work resulted in 5 publications in conferences such as KDD, SIAM SDM and ASONAM.

Livermore, Ca, Us

• Designed and tested a variety of time series classification models (k-NN, DTW, SVMs, CNNs, autoencoders) to identify computing failures in multi-physics simulations.• Used time series interpolation and augmentation methods to address varied lengths in data instances and to improve class imbalance.

Albuquerque, Nm, Us

Transferred to UC Riverside following my PhD advisor.

Praha, Cz

Doctor Of Philosophy - Phd, Computer Science

Master Of Technology - Mtech, Computer Science

Bachelor Of Technology - Btech, Computer Science