Joseph Thangraj, Ph.D.

I successfully led initiatives to build advanced GenAI solutions that transformed healthcare documentation processes. • Developed robust data pipelines that improved data handling and processing efficiency. • Championed the integration of AI tools that reduced document generation time. • Fostered collaboration among diverse teams to ensure alignment with strategic healthcare goals, driving innovation.

Clustering ambient seismic noise to identify seismic noise sources using Deep Clustering with Variational Autoencoders - Baylor University, 2020 o Engineered spectrograms from continuous seismic noise as features. o Used variational autoencoders to reduce the dimensionality of the data. o Clustered data to separate oil pump jack, vibroseis and incoherent sources.

I strive to improve student involvement and help organize various events for Baylor Graduate student Association. In addition to that, I manage various social media communication channels like slack, email, etc to keep all the representatives in the team connected and updated.

Designed an IoT-enabled solution for real-time acquisition, transmission, and processing of seismic time-series data, specifically targeting low-cost geothermal studies. - https://doi-org.ezproxy.baylor.edu/10.1016/j.jappgeo.2021.104426) o Collaborated with the Department of Energy to implement continuous time-series data usage, advancing geothermal exploration. o Created an efficient ETL pipeline using Apache Cassandra for storing and retrieving time series data. o Developed a Python-based data processing pipeline to extract geothermal signals from continuously recorded data. o Leveraged SQL for bookkeeping and experiment tracking. o Engineered ML models by extracting time and spectral features from seismic signals.

Summer Research Internship - Machine learning, University of Texas Austin, Prof. Mrinal Sen, 2021 o Using K-means for unsupervised clustering of ambient noise seismic data (time series data) to determine high-quality seismic signals for subsurface modeling (under peer-review) o Engineered features from continuous time-series seismic data using cross-correlation and spectral analysis (CUDA and tensorflow) to get information on seismic source locations, direction and velocity o Developed tools in python to calculate frequency, amplitude and source location features. o Smoothed and down sampled the features to reduce computation cost. o Used unsupervised clustering to find coherent seismic sources associated with water injection.

Developed a plug-in for Petrel in Ocean environment(Gas Hydrates Interpretation Tool) under the Schlumberger Ocean Competition 4. (Reached the finals among 110 entries)

Seismic Interpretation(Leads were Identified). Original Oil in Place (OOIP) estimation and geological probability of success OOIP estimation using deterministic and Monte-Carlo probabilistic methods

Created a data-driven solution to create synthetic seismograms for subsurface modeling

Core cutting and Sampling. Multibeam Bathymetry, Magnetic methods, sub-bottom profiling and speed velocity methods.

My projects involve the implementation of machine learning and deep learning algorithms to solve time series and ambient seismic noise research problems