AI Solutions

Wealth Management Reporting & Analytics

Wealth Management Reporting & Analytics- Collaborated with cross-functional teams to implement real-time data ingestion pipelines using Python and Logstash, enabling Trade data streaming every 5 minutes between AWS S3, On-premises S3, and Elasticsearch- Analyzed and mapped service level dependency data from Dynatrace to Neo4j using Java Spring Boot, leading to robust data monitoring on the health status of affected services and reducing incident response time- Collaborated with Data Analysts to train machine learning models using Python, conducting an in-depth analysis of the factors influencing critical status among various applications, which facilitated informed decision-making- Conducted a comprehensive proof-of-concept study of Amazon SageMaker, leveraging various machine learning models to explore data insights and present findings to the team- Created watches in Elasticsearch that trigger alerts in real-time to MoogSoft for anomaly detection to reduce incident response time- Created Kibana dashboard utilizing aggregations for Kafka metrics

Supervised by Prof. Igor Gilitschenski

Supervised by Prof. David Lindell and Prof. Kyros Kutulakos- Captured compressive snapshot images with coded camera sensors, using custom exposure patterns and different mask sizes to enable dual-tap pixel-level programmable control over image acquisition - Implemented and trained deep learning models, including fully-connected networks and U-Nets proposed in various research papers, to achieve high-speed video reconstructions - Integrated Fast Forward Computer Vision to accelerate data loading and model training, resulting in 1.5x speedup

Supervised by Prof. Bo Wang- Standardized medical image data format and generated paired bimodal images from mouse PET/CT scans and contours at different timepoints after nanomedicine injection to analyze nanomedicine delivery for cancer treatment - Trained 3D U-Net models with 5-fold cross-validation using the SOTA image segmentation method, nnU-Net, achieving automatic organ and tumor quantification with maximum 98% volumetric accuracy - Summarized and submitted the final results as a paper to multiple conferences

- Designed and implemented the user interfaces for a web application with Django, using JavaScript, HTML and CSS- Achieved users’ access control to resources according to roles by designing the MySQL database and URL rules- Implemented an employee salary management system, which allowed users to execute basic SQL commands to manipulate salary data, and display statistical and graphical summaries, to help the company manage employees