Experience

Armorblox (now part of Cisco)

• United States
• Software Development
• 1 - 100 Employee

• Senior Machine Learning Engineer

  • Oct 2022 - Present

  Natural Language Understanding applied to Cybersecurity Natural Language Understanding applied to Cybersecurity



Stovell AI Systems

• United States
• Information Technology & Services
• 1 - 100 Employee

• Senior Staff Machine Learning Engineer

  • Aug 2021 - Sep 2022

  Successfully developed a Transformer model for a time-series forecasting application with a direct impact on company revenue, leveraging state-of-the-art machine-learning techniques. Built a quality, product-agnostic platform for distributed, data-parallel training of PyTorch models with Ray Train and Ray Tune on a large GPU cluster. Built a quality, product-agnostic data-engineering system atop Ray Core to convert raw, tabular data into an ingestible dataset with proprietary engineered features on a large CPU cluster. Exploratory analysis of raw data with Jupyter notebooks and Matplotlib.



Change Healthcare

• United States
• IT Services and IT Consulting
• 700 & Above Employee

• Senior Machine Learning Engineer

  • Aug 2020 - Jul 2021

  I worked on a team called Coordination of Benefits which performs fuzzy matching between demographic records in two separate datasets: a client dataset and a trading-partner dataset. I was responsible for performing data engineering on the trading-partner dataset, which is in the neighborhood of a terabyte in size. The preprocessing is comprised of several stages, including hashing columns, merging with older versions of the dataset, slicing the demographic columns, and predicate-based partitioning for efficient comparisons. Microservices built around PySpark were used to deploy the data preprocessor. The overall service is triggered when a new client or trading-partner file is uploaded. Each file is automatically preprocessed and compared with the opposing dataset for fuzzy matches. The model is a logistic regressor trained on synthetic and real records. I also built a service which applies transformations to an input dataframe via PySpark. This general-purpose service has been used across several projects within the AI org.



Abacus.AI

• United States
• Software Development
• 1 - 100 Employee

• Machine Learning Engineer

  • Jan 2020 - May 2020

  Given only a dataset and a minimal amount of metadata, our AutoML platform discovers and trains a Deep Learning model that can solve a variety of problem types, including forecasting, anomaly detection, regression, recommendation, and personalization. I have implemented features that span the entire AutoML process, including efficiently preprocessing data with PySpark, designing efficient and practical AutoML algorithms, exposing new inference capabilities to our public API. Given only a dataset and a minimal amount of metadata, our AutoML platform discovers and trains a Deep Learning model that can solve a variety of problem types, including forecasting, anomaly detection, regression, recommendation, and personalization. I have implemented features that span the entire AutoML process, including efficiently preprocessing data with PySpark, designing efficient and practical AutoML algorithms, exposing new inference capabilities to our public API.



Amazon Web Services (AWS)

• United States
• IT Services and IT Consulting
• 700 & Above Employee

• Deep Learning Backend Engine Developer

  • May 2018 - Sep 2019

  Hardware-centric AutoML research. Wrote a parser to optimize inference latency of trained PyTorch models for deployment. Designed a nightly test harness to monitor performance of SageMaker Neo. Hardware-centric AutoML research. Wrote a parser to optimize inference latency of trained PyTorch models for deployment. Designed a nightly test harness to monitor performance of SageMaker Neo.



Symbio Robotics

• United States
• Automation Machinery Manufacturing
• 1 - 100 Employee

• Machine Learning Engineer

  • 2016 - May 2018

  Based in the San Francisco Bay Area, Symbio is dedicated to strengthening Industrial Automation Systems through the development of Robotics and Artificial Intelligence technologies. Shortly after graduating from UC Berkeley’s incubator program in 2015, Symbio structured contracts with the Ford Motor Company and quickly gained support from private and federal funding sources. Successfully applied a neural network-based Reinforcement-Learning algorithm called Guided Policy Search to a six-dimensional industrial automation task. Prototyped, implemented, and deployed a Reinforcement-Learning algorithm based on Bayesian Optimization to estimate parameters of robotic motion primitives. Prototyped, implemented, and deployed a dynamical robotics model based on Gaussian Mixture Models.