Speech team - keyword spotting. Process audio data, data cleaning, featurizing data, AI models for edge devices. Essentially, I'm part of the team that makes "Hey Google" work.

• Partnered with 2 traditionally non-tech companies in the medical domain to build and deployend to end machine learning (ML) and data pipelines from the ground up.• Designed a framework for solving differential equations with deep learning in pytorch, whichhad a 10-100x speeedup compared with traditional gradient descent methods.• Created explainable ML models using SHAP analysis to determine why users were churning.• Used a distributed system architecture to create an email scheduling and sending system, withDynamoDB in the backend to provide instant lookup of 10 million user records.• Built a searchable database of leads for the sales team at a pharmaceutical company, usingAirflow to orchestrate the ingestion of 5+ data sources into a single ElasticSearch database.

• Advised two insight fellows on how to quickly and effectively implement machine learning models in their four-week Insight project.• Provided mock behavioral and technical interviews for fellows to help them prepare for interviews with connected companies.

• Used computer vision methods to automatically generate videogame highlights to aid ESports marketing.• Incorporated a pre-trained CNN (vggish from Tensorflow) to convert audio clips into embeddings.• Implemented a classifier using transfer learning in PyTorch by training a classification layer on vggish determined embeddings.• Designed a pipeline that automatically generates suggested labels, trains, and evaluates different models.• Partnered with an early stage startup, Visor, to integrate the model into their code base.

• Created new open-source Python packages to learn protein model parameters from experimental data, and contributed ~90% of the code while maintaining the package for the research group.• Identified and selected ideal features in protein simulations using dimensional reduction (PCA).• Used non-linear optimizers (i.e. stochastic gradient descent, L-BFGS-B) to train protein models.• Optimized package by parallelizing >75% of the computation time using MPI (mpi4py), and profiled memory and processor utilization to determine the slowest sections of the code.• Managed a team of 2 undergraduate researchers and mentored them on research projects.

• Discovered a new material that can be chemically tuned between an semiconductor and semimetal.• Used Matlab and Origin to perform numerical analysis on resistance vs temperature plots to determine the semiconductor band-gap.• Mentored and trained 3 undergraduate researchers in a lab environment and guided them in completing experiments and presenting their work.• Analyzed materials using x-ray and neutron crystallography, magnetic, and resistance measurements.• Operated and maintained high vacuum and low temperature measurement equipment.• Handled scheduling and allocation of lab instrument time between 6 group members.

• Taught a bi-weekly 3 hour long introductory physics laboratory sessions for approx. 30 students.• Assisted students in completing experiments and instructed on how to report results appropriately.

• Volunteered with PAGSA to improve professional opportunities for graduate students at Rice University.• Organized networking and social events for the approximately 100 students in the department.• Participated in monthly meetings to approve grant funding (up to $1,000 per) for school organizations.

Tested graphene window flow properties for use as a UV window in vacuum systems. Designed and built instruments for the lab's Angle Resolved PhotoEmission Spectroscopy (ARPES) system.

Studied ultra-thin Germanium films for their use in photovoltaic cells. Designed, built and calibrated a Hall-effect measurement setup.