Ads ranking - CoreML - ML Foundation and Model System Co-DesignPushing the boundaries of ads modeling by advancing the SOTA of deep learning in architecture design, training algorithms, efficient ML/DL techniques (e.g. model compression, knowledge distillation, dynamic architecture), data/model parallelism, multi-task learning, model-system co-design, etc.

Lead design and execution of a data startup focused on translating the digital RF world into the physical world. Creating the next generation of security via decomposing cellular, wifi and bluetooth signals around us.- Engineer number 1; Created initial prototypes- Responsible for several generation iterations on our cloud architecture, applied machine learning research, backend software, security, dev ops CI/CD flows, and engineering operations.- Grew several teams simultaneously relating to software, infrastructure, research, and external API integrations. Responsible for an org of 14 engineers across 3 core teams.- Created novel approaches for collecting and building models on top of Cellular, Wifi, and Bluetooth data streams to classify the ever moving targets of real world electronics.- Continuous cost optimization applied to evolving cloud design that ingests extremely high velocity data- Designed and built internal machine learning platform and stream computation engine that focuses on key principals of realtime data, historical time series, easily deployable offline models, massive simulation, and visualization. Inspiration taken from past DARPA programs focused on AutoML, active learning, along with ingenuity dealing with requirements of high velocity data and realtime consumption.

Some high level, but not all inclusive, work I have been involved in at JPL:- Working to improve the Deep Space Network scheduling process via reinforcement learning- Uncertainty quantification surrounding radiation effects affecting performance of electronic components- DARPA D3M - Helped design and implement facets of an AutoML system (https://www.darpa.mil/program/data-driven-discovery-of-models) My work focused on designing and implementing a neural architecture search framework- DARPA LwLL - Designing and implementing the evaluation framework and API for top deep learning researchers across the world who are pushing the state of the art in active learning, few shot learning, and transfer learning as it relates to large unstructured data problems within image classification / object detection / machine translation (https://www.darpa.mil/program/learning-with-less-labels)- Building an automated anomaly detection framework for spacecraft telemetry data using dynamic thresholds and online deep learning models robust to domain shift- Computer vision models for minimizing energy consumption of planetary rovers- Image segmentation of electronic components for auto determining risk in radiation effects for space hardware- Anomaly detection operations software for the Voyager mission, which are the only man made objects in interstellar space

Managed 32 students who plan and put on events at Augustana College. Allocated budget to 13 different committees along with other office expenses.

Created a simulation for wave propagation. Simulated single photon sources for their use in quantum networking as well as developing undergraduate labs for physics majors.

Worked with clinical radiation data. Presented research at an MD Anderson monthly radiation oncologist meeting.

Conducted data analysis and programmed to analyze data for neutron rich isotope experiment. Funded to present research at the national American Physical Society conference.