Professional Experience

Capital One, McLean, VA

Guardrails for LLMs: Designed and implemented safety measures and guardrails for large language models, ensuring robust and secure AI interactions.
Search Functionality in EASE Mobile App: Enhanced user experience by improving search efficiency and effectiveness in the help page.
Transformer-based Autocomplete Feature: Improved user interaction and search functionality within the EASE mobile application.

Dialog Manager for Chatbot Eno: Led the design and deployment of a new dialog manager using Rasa and transformer models to handle multi-turn conversations, significantly improving performance metrics.
Conversation Grading Model: Developed a transformer-based model to establish a robust performance metric for evaluating conversational AI systems at Capital One.

Predictive Modeling: Implemented a CNN architecture using Keras and TensorFlow to forecast rain fades of satellite signals using GOES-16 images, achieving a 90% F1-score.
Dynamic Inroute Association: Developed a reinforcement learning algorithm to dynamically allocate inroute links to terminals in a satellite network, maximizing system throughput.
Dynamic Resource Allocation: Created an LSTM model to predict future traffic characteristics of a satellite communications network.

Deep Learning: Implemented a neural network architecture using Keras that improved the win-rate predictor accuracy at Carvana by 15%. The predictor estimates Carvana's probability of winning a vehicle at auction based on bid price and other factors.
Data Visualization: Created a Tableau dashboard for data reporting, enabling effective comparison of 2018 and 2019 vehicle auction sales.

Vehicle Network Optimization: Developed an optimized control and learning system to improve traffic flow and reduce accident risk in autonomous vehicle networks.
Cyber-Physical Security: Enhanced the security of autonomous connected vehicle networks against cyber and physical attacks.
IoT Signal Authentication: Designed a dynamic IoT signal authentication algorithm using Long Short-Term Memory (LSTM) models.
Distributed Data Generation: Developed a Generative Adversarial Network (GAN) framework for distributed and privacy-preserving data generation.
UAV Wireless Connectivity: Explored machine learning techniques for enhancing wireless connectivity for UAVs.
UAV Security: Worked on security measures for cellular-connected UAVs.

Android Application: Designed, developed, and implemented an Android application in Java for real-time audio signal processing.
iOS Application: Developed the back-end process for a multimedia iOS application in C.