Vincent Filardi

Introduces an economical, non-intrusive methodology for precision subsurface characterization, leveraging advanced DL and ML techniques to accurately estimate soil water content at various depths. Gen-AI solution trained on unstructured data to emulate realistic labels for transfer learning on a small dataset. Engineered and extracted the most relevant signal features associated to soil water content via unsupervised DL to achieve at most 6% mean absolute error across multiple soil depths.

Time series deep representation learning applied to intradialytic chronic kidney disease patient data. Processed 128TB of 4.8 million hemodialysis sessions using PySpark and AWS Glue. Collaborated on RAG solution on top of Gemini and Med-PaLM for querying corpus of technical documents. Engaged with clinical experts to create impactful visualizations in R, highlighting patterns in 4.8 million hemodialysis sessions, leading to 3 publications in the Journal of the American Society of Nephrology.

NSF grant CCF-210452.

DS 1010: Data Science I: Introduction to Data Science DS 1020: Data Science II: Modeling and Data Analysis

My team and I worked on the Renal Research Institute's problem of pre-processing and model identification from sparse non-uniform data. The Math Problems in Industry (MPI) Workshop is a lively, 5-day interaction on problems of interest to science and industry.

This research is in partnership with NOAA CREST, the Computer Science Department and the Mechanical Engineering Department of CCNY. Used multidimensional projection techniques on a high-resolution urban weather and landcover dataset to characterize and visualize the intercity variability in critical climate variables and the key factors that influence them. We use semi-supervised and unsupervised learning techniques to discover relationships and tune the model. Techniques: Kernal PCA, Least Squares Projection, Isomap, Canonical Correlation Analysis, Image Decomposition, Parallel Coordinates, Pixel Index Map Advisors: Professor Ronak Etemadpour & Professor Prathap Ramamurthy I received additional funding from the City College Fellowship in support of this project.

Hired to conduct Fellowship sponsored research under a mentor at CCNY, presented at major conferences.

Responsible for homework and exam grades of 40 students and provided feedback on work.

This research was in partnership with GE Aviation, NSF, Center for Industrial Mathematics and Statistics of WPI. Constructed models to predict material wear, determine the most effective material combinations and extract the features of the data that were most critical to the overall wear. Improved predictive capabilities on sub 150-row datasets by up to 300% by performing unsupervised dimensionality reduction and using ensemble methods for prediction. Presented internally at GE Aviation headquarters in Cincinnati and 2 academic conferences. Techniques: Kernel PCA, Autoencoders, Regression, Boosted Random Forest, Cross-Validation. Advisors: Professor Randy Paffenroth, Professor Burt Tilley & Dr. Ming Fu I received the City College Fellowship Summer Research Grant and Dr. Barnett and Jean Hollander Rich Research Experience Scholarship in support of my work on the project.

With little data science experience constructed and models to predict material wear, determine the most effective material combinations, and extract the features of the data that were most critical to the overall wear. Improved predictive capabilities on sub-150-row datasets by up to 300% to achieve industry standards. Presented internally at GE Aviation headquarters in Cincinnati and two academic conferences.

This research was in partnership with the NSA and Mathematics Department of Miami University. Modified the Kermack-McKendrick model used to simulate an epidemic outbreak. Introduced new non-linear terms and parameters into the system of PDEs encapsulating the effect of prevention policies and the spatial transmission of disease. Presented to the public and faculty at Miami University and an academic conference. Advisor: Professor Alin Pogan Techniques: Spectral Analysis, Phase Plane Analysis, Newtons Method, Runge-Kutta Method, Nonlinear Analysis.

PhD Advisor: Professor Oren Mangoubi Research Areas: Subsurface imaging

Advisor: Oren Mangoubi Research Areas: Optimization, Meta-Learning, Adversarial Learning