Senior Machine Learning Engineer, Senior Data Scientist2019 - PRESENTSelf-employed
Technologies: TensorFlow, OpenCV, Deep Learning, Natural Language Processing (NLP), Machine Learning, Python
- Developed an end-to-end REST application for machine learning model management and microservices using MLflow, including fastText, and BERT encoders.
- Developed state-of-the-art NLP models in multiple languages for named entity recognition, sentiment analysis, topic analysis, and intent recognition using TensorFlow, Keras, and SpaCy.
- Created an online learning process to continuously train algorithms, reduce annotation labor costs up to 30%, and integrate into a chatbot application.
- Developed customer segmentation classifiers based on sales, orders, and customer demographics to drive revenue increases by ~20% for a retail company. Applied computer vision to detect and annotate products automatically based on produce images.
- Worked on developing an options trading algorithm for the client to maximize profits in volatile markets and detect underpriced assets. This is an ongoing process.
- Created predictive maintenance and wind power optimization models in Python for wind turbines. Architected a cloud solution and automated a reporting pipeline using Spark, Databricks, Azure, and SQL.
- Developed cutting-edge real-time computer vision models (Deep Sort, YOLO, and Siamese networks) for multi-object tracking and object detection on drones. Created an automated pipeline for training, saving up new detection categories' costs to 90%.
Senior Data Science Consultant2019 - 2020Guidehouse
Technologies: SQL, Spark, R, Python
- Led a utility asset risk model development to inform over USD $5 billion in grid resiliency planning. This included an ETL pipeline for hundreds of files and formats from scratch, weather modeling, classifiers for imputation, and graph analysis.
- Managed a USD $150,000 machine learning project to predict auto accidents and grid impacts for an electric utility. Collected roads, weather, utility, and traffic data. Built several classifiers with an AUC ROC of 89%. Led client technical workshops.
- Led big data proof-of-concept using Spark, R, and Scala to process TB-sized data sets. This included a robust standard error package in Spark, dynamic time warping machine learning tools for customer segmentation, and Spark transformation pipelines.
- Spearheaded the development of an internal weather package to process NOAA data for firm-wide projects. Created FlexDash and Shiny dashboards, as well as parameterized QC memos to visualize data and assess completeness.
- Served as a lead modeler of the Bass diffusion model for forecasting electric vehicle (EV) adoption and EV siting analysis. Applied linear programming and optimization techniques to improve analysis times over 200%.
- Created a model to predict window stock, turnover, and efficiency changes in the United States using Bayesian inference with Dirichlet priors. The automated approach allowed us to decrease client costs for the project by ~40%.
Data Engineer2019 - 2019Hays Consulting
- Architected the enterprise ETL solution to extract data from data lakes and major ERPs, process it using ephemeral MemSQL clusters, and update data warehouses. Included REST APIs, Airflow, and dynamic SQL.
- Developed a custom QC and testing suite in Python to perform regression, integration, and unit testing. Quality checks and Type 2 tracking ensured the highest data integrity.
- Developed process mining and outlier analysis tools including custom dashboards using D3 and Zoom.
Software Engineer2018 - 2018Payger
Technologies: Amazon Web Services (AWS), ELK (Elastic Stack), BitShares, EOS, AWS, Java
- Developed a blockchain payments platform on the BitShares network to reduce settlement times by over 1,000% compared to traditional methods.
- Designed an Elasticsearch back end and micro-service architecture using Java and AWS for data management and processing.
- Delivered a block explorer REST application for real-time transaction monitoring of the BitShares network.
Professional Research Assistant2014 - 2016Laboratory for Atmospheric and Space Physics
Technologies: Swift, IDL
- Developed lunar dust and mass spectrometer models to process millions of image-charge signals for the LADEE lunar mission. Presented work at AGU 2015.
- Developed image processing tools in IDL and Swift for dust accelerator calibration experiments.
- Led development of the SUDA mass spectrometer lab prototype for Europa's Clipper mission. Worked across science, engineering, and simulation groups to fabricate mechanical and electrical components and construct a working device for under $10,000.