AR Travel Planning App: Overcoming Bottlenecks by Balancing Storage, Computation, and User Experience on Mobile Devices

This project aims to develop an Augmented Reality (AR) travel planning app that integrates side quests and in-game advertising. The goal is to provide a smooth and immersive AR experience on a wide range of smartphones, ensuring accessibility for both high-end and lower-end devices. We focus on balancing performance demands with user experience to create a seamless and engaging app.

AR Travel Planning App: Uncertainty Planning through Mobile Telemetry

Automated Feature Engineering Tool for Heart-Brain Dynamics

The project applies deep learning techniques for causal discovery in ECG and EEG signals. Causal discovery is the process of inferring causal relationships from data. This is in contrast to traditional statistical techniques, which often focus on correlation. Causal discovery aims to identify whether one variable directly affects another. For this study, we use deep learning techniques to identify causal relationships between ECG, and 6 different channels for EEG - and how these relationships change during the different stages of sleep (wakefulness, stage 1, stage 2, stage 3, REM). By improving our understanding of these dynamic relationships, we can potentially open new avenues for therapeutic interventions in sleep disorders where heart-brain interactions play a pivotal role.

Path-Based Abduction Learning for Pruning Exponentially Large Search Trees

Path-based abduction learning leverages the principles of abstraction interpolation and constraint programming to generate generalized and useful interpolants to effectively reduce the search space in combinatorial search problems.

LEVERAGING MACHINE LEARNING AND LEARNED BLOOM FILTER FOR MALICIOUS URL DETECTION

As online usage increases, cybercriminals exploit malicious URLs to target users, resulting in financial theft, identity fraud, and malware installations, with annual losses in the billions’ of dollars. As modern applications generate more data, the demand for scalable and eLicient methods for storing, retrieving, and verifying this information grows. As a result, improving methods for detecting malicious URLs has become critical for cybersecurity. Here Machine Learning and Learned Bloom Filter is used to lower the FPR and False classification rate.

Machine Learning at the Network Edge

Combining machine learning with bloom filters to create a memory efficient data structure for storing classification data

Towards Fine-grained Breast Cancer Risk Assessment Using Large Multimodal Model

This project represents the first attempt in utilizing large multimodal model for breast cancer risk assessment.

OfferAI - Job search engine in the AI era.

"Specifically designed for international students' job hunting, addressing three core pain points: 1.Not knowing how to revise the resume – Trained a resume revision model based on the expertise of professional HR. 2.Uncertain about which positions are suitable – Developed a job matching model incorporating insights from professional career consultants. 3.Unsure how to prepare – Created a job search AI engine that helps users instantly find the latest and most comprehensive information and provides answers to their questions."

Agent4Rec: AI-Powered Recommendation Simulation Platform

Agent4Rec offers faithful and effective simulated user behaviors using LLM-driven generative agents, enabling convenient testing and optimization of recommendation systems.

Generative AI Accelerators to improve cost savings and CSAT in Airline Operations

Projects focus on improving operational efficiency and enhancing customer satisfaction (CSAT) in airlines using cutting-edge Generative AI technology. We’ve developed two key solutions: E-Validator and Engineer Assist, both designed to drive cost savings and streamline operations.

Optimizing LLM Chatbots For Business Use Cases

The development of LLM-based chatbots is transforming how businesses handle customer interactions, information retrieval, and process automation. This project focuses on optimizing Large Language Model (LLM) chatbots for complex business use cases. By leveraging cutting-edge techniques such as Retrieval-Augmented Generation (RAG) and entity extraction, this project aims to enhance the scalability, accuracy, and adaptability of AI-driven chatbots within diverse business environments.