Built data pipeline to process time-series concentration data collected by IoT sensors at industrial chemical plants.

Designed and implemented a dynamic algorithm to correct sensor drift and identify anomalies in the spectral data.

Developed script to interact with client data in Cloud Firestore and automate the data modeling process, reducing the analysis time by 90%.

Built multi-output regression and classification neural networks to predict species concentrations for various chemical plants, reducing industrial operating costs by ~$50K.

Designed (CAD) and installed a biodiesel vector system onto a diesel-powered tractor for usage on the university farms to analyze the benefits of switching to B100 biodiesel.

Used data science tools such as Pandas and Keras to interpret biodiesel operating data and quantify the effectiveness of the vector system technology.

Performed a life cycle analysis of the emissions impact of switching campus vehicles to B100 biodiesel.

Researched and developed deep learning models based on BERT and Detectron to evaluate visual question-answering (VQA) tasks on seismic images.

Designed a proprietary framework to train deep learning model on ~10,000 seismic images via transfer learning.

Built and presented a proof-of-concept VQA model with the potential to speed up seismic interpretation process 10x and reduce costs of exploratory drilling by ~$50M.

Designed and developed photosensors to measure water turbidity for sustainable water treatment research.

Programmed microcontrollers (Arduino) to read sensor output and convert to calibrated concentration measurements.

Created a mobile app that connects to a borescope and converts image data to calibrated turbidity readings.