Suhas Cristy Mathey

Graduate Research Assistant

• Worked on an inter-disciplinary project involving hyperspectral Image Analysis using unsupervised learning techniques.
• Utilized matrix decomposition techniques from linear algebra to handle high-dimensional image data, transforming it into lower-dimensional subspaces for efficient Image analysis & Processing
• Developed and implemented machine learning algorithms in Julia to cluster and analyze hyperspectral image datasets using unsupervised learning methods
• Presented research findings through weekly lab meetings using data visualizations created with CairoMakie, focusing on clear and actionable insights from hyperspectral data analysis
• Utilized GitHub for version control and collaboration with the research team to manage the development workflow effectively
• Developed a Julia software package, SubspaceClustering, which implements advanced subspace-based algorithms for efficient clustering and analysis of high-dimensional data

Computational Researcher

• Implemented advanced numerical computations in Python using CuPy to leverage GPU acceleration
• Developed and manipulated synthetic datasets using Pandas, enhancing ML models for accurate prediction of atomic scattering patterns
• Engineered parallel processing solutions via Thread Pool Executors for computational efficiency

Senior Multimedia Student Assistant

• Managed a team of 8 to help students and faculty address their multimedia needs
• Collaborate with individuals from diverse backgrounds and effectively communicating in utilizing the multimedia resources and services
• Worked on Video Editing project using Final Cut Pro on archived footage of University’s first Study Abroad Trip
• Awarded the 2024 Library Student Assistant Scholarship award in recognition of my work and demonstrated leadership potential

IT Field Support Technician

• Provided IT support to the University’s facilities department, comprising about 500-600 employees, and managed an average of 1300 client devices including iPhones, iPads, and Laptops
• Configured local system network and installed security certificates to route client’s devices to the university Wi-Fi
• Imaged laptops and desktops to the client's needs through MDT and deployed them in their workspace
• Maintained up-to-date customer records on Team Dynamix and interacted with customers through email, phone, and in person
• Addressed domain problems on the university-owned laptops and managed the devices through Active Directory
• Used TeamViewer to remote into clients’ computers for troubleshooting purposes

University of Delaware

GPA: 3.67

Indian Institute of Information Technology, Tiruchirappalli

GPA: 3.1

Pigment Mapping in Reflectance Imaging Spectroscopy (RIS) Datasets - Unpservised Learning

• Developed and optimized machine learning algorithms in Julia for pigment mapping, to enhance clustering performance and accuracy in hyperspectral RIS datasets.
• Implemented clustering techniques to group pixels based on spectral signatures, aiding in the identification of pigments.
• Achieved over 95% accuracy in clustering results through the development and optimization of machine learning algorithms for spectral signature-based pigment identification in hyperspectral imaging datasets.
• Presented clustering analysis and research findings at technical conferences, including a keynote talk at Techno Heritage 2024 and a poster session at the 2024 IEEE Baltimore Technical Colloquium, underlining the contribution to technical innovation in the field of Machine Learning and signal processing.

Biomolecular Identification Algorithm for Mass Spectrometry

• Investigated bioinformatics methods for identifying several biological targets with minimal spectrum pre-processing needed.
• Implemented and compared clustering algorithms, including K-Means, Spectral Clustering, and K-Subspaces Models, to enhance the identification and classification of biological targets.
• Designed a robust pipeline for data preprocessing, dimensionality reduction, and clustering, optimizing the algorithm’s accuracy and performance on bioinformatics datasets.
• Validated the algorithm's effectiveness on diverse datasets, demonstrating its adaptability and reliability in identifying biomolecular targets across varying experimental conditions.

Hyperspectral Image Analysis - Unsupervised Learning

• Implemented clustering algorithms, including K-Means, Spectral Clustering, and K-Subspaces Clustering, for hyperspectral image analysis.
• Analyzed and classified hyperspectral datasets such as Pavia, Salinas, Indian Pines, and Onera Satellite, focusing on spectral and spatial feature clustering.
• Optimized clustering techniques to enhance accuracy in segmenting high-dimensional hyperspectral data.
• Developed scalable workflows to process and visualize results, aiding in comprehensive analysis and actionable insights from hyperspectral imagery.

Feature Engineering with Analytical Data - Dow Chemical

• Implemented data driven statistical modeling on unstructured spectroscopy datasets namely, FTIR, GPC, and NMR to develop a quantitative polymer structure - performance relationship
• Employed different data pre-processing techniques like Binning, Peak Normalization, and PCA
• Data Pre-Processing techniques are used for improving accuracy of the spectral analysis and reducing dimensionality for data modeling
• Implemented linear regression models in analyzing the relationship between polymer structure and performance
• Utilized Matplotlib for comprehensive data visualization, including scree plots, parity plots, and other graphical representations

Movie Genre Prediction - NLP

• Developed a movie genre prediction model using IMDb dataset with 10,000 movies by preprocessing text descriptions through stop word removal and lemmatization
• Employed BERT Tokenizer for text tokenization and utilized GloVe embeddings to construct a comprehensive word embeddings matrix
• Designed and trained an LSTM model, leveraging word embeddings for feature extraction, to accurately classify movie genres based on descriptions
• Currently working on improving the performance of the model

Image Super Resolution using Deep Learning

• Used SR Generative Adversarial Network (GAN) to restore the quality of low-resolution images
• Applied a pre-trained Convolution Neural Network (CNN) to improve the quality of a digitally zoomed image
• Created a dataset of 20,000 images to train the network and 10,000 images to test the network
• Increased the resolution of an image by 30%

Body Posture Detection using Mediapipe Framework

• Built an ML pipeline using MediaPipe Framework to detect wrong postures while exercising
• Used SciPy Library to calculate the cosine similarity between the correct form and the wrong form of exercise
• Utilized postures from Cornell’s Physical Therapy database as our benchmarks for form comparison
• Implemented the model for a safer and more efficient training regimen for athletes while working out

HOPE WorldWide Ltd

• Active volunteer in Hope Worldwide since freshman year in college
• Organized a volunteer event to renovate Hope Worldwide orphanage in Tiruchirappalli, India with 200+ participants
• Organized a clean-up with about 50 people from the locality at White Clay Creek Park in Newark, DE

Hen Hacks, University of Delaware

• Participated in a 24-hour hackathon conducted by the University of Delaware
• Built a website in improving financial literacy in underserved communities in Delaware
• Successfully launched a website to facilitate the community in providing the best resources when it comes to looking out for a loan

2024 Library Student Assistant Scholarship Award

Since 1996, the Library, Museums and Press Student Assistant Scholarship rewards outstanding student service at the University of Delaware Library, Museums and Press. The scholarship program serves as a means of recognizing outstanding student empoloyee's work and leadership potential, as well as supporting their pursuits of a graduate education