Data Scientist | Machine Learning Engineer Contractor
2019 - PRESENTParabolica Labs- Provided machine learning and data science solutions with a specialization in natural language processing (NLP), time series analysis problems, applications of machine learning to sensor data, classical machine learning algorithms, deep learning models, Bayesian statistical models, and data engineering tasks on large scale data.
Technologies: Python, Cython, SQL, NumPy, Pandas, Scikit-learn, SciPy, AWS, Git, Hadoop, Hive, Impala, Databricks, Dask, SpaCy, NLTK, XGBoostMachine Learning Engineer
2018 - 2019Twosense, Inc.- Researched, developed, and deployed a suite of machine learning models (NumPy, SciPy, scikit-learn, XGBoost, Cython) that authenticated users based on behavioral biometrics collected from sensors on phones and computers.
- Wrote large-scale data processing scripts that consumed real-time biometric data for model training and testing using NumPy, AWS Redshift, and Pandas.
- Produced Jupyter notebooks visualizing model validation metrics, data transformations, and feature analysis.
- Guided best practices, led technical sessions, collaborated on project specifications, and wrote significant amounts of research documentation. I was hired as the first member of the machine learning and data science team.
- Wrote suites of unit tests for data processing, feature, extraction, and model validation.
Technologies: Python, Cython, SQL, NumPy, Pandas, Scikit-learn, SciPy, AWS, GitSenior Data Scientist
2017 - 2018Skedaddle- Developed a production-grade API for pricing algorithms using NumPy, scikit-learn, Lambda, and API Gateway.
- Built and maintained complete data pipeline platform reading public APIs using EC2, Lambda, and Snowflake, for creating time series models predicting product demand.
- Wrote a serverless web app displaying data visualizations and real-time monitors of key metrics using Flask, Zappa, and D3.js.
- Provided ad hoc analysis for all domains within the organization, and guided other team members in their analyses.
Technologies: Python, Flask, SQL, Snowflake, Pandas, NumPy, Scikit-learn, SciPySenior Data Scientist
2016 - 2017Whoop, Inc.- Led team code reviews and collaborated on the direction of quarterly team goals and projects.
- Researched, developed, and deployed novel, deep, convolutional neural networks for classifying activities based on multidimensional sensor data using PyTorch.
- Developed and maintained data pipelines consuming from Redshift and PostgreSQL databases.
- Built a real-time activity detection algorithm for biometric time series data using NumPy and SciPy.
- Researched and developed convolutional autoencoder models for compressive sensing using PyTorch.
- Wrote a real-time algorithm to detect how the user wears a sensor packet based on accelerometer and biometric profiles using NumPy, SciPy, and scikit-learn.
Technologies: Python, SQL, PyTorch, TensorFlow, SQL, AWS, Redshift, Scikit-learn, Pandas, NumPy, SciPyAssociate Data Scientist
2014 - 2016Cogo Labs- Implemented a Python library for A/B testing with Bayesian Statistics and other measurement tools.
- Developed statistical methods to mine URLs from user clickstream data (Presto) and developed neural networks (Tensorflow, Keras) to model user level characteristics from the browsing history.
- Applied NLP (scikit-learn, NumPy) techniques to cluster ad campaigns based on content similarity.
- Wrote algorithms (NumPy, SciPy, Pandas, scikit-learn) for user-campaign selection and developed models based on clickstream data, market intent, and demographics.
- Built tools to monitor critical metrics and score production models daily using Python, MySQL, PostgreSQL Presto, and MapReduce.
Technologies: Python, SQL, TensforFlow, Keras, Pandas, SciPy, NumPy, Scikit-learnFull-stack Developer
2013 - 2014Microsoft Project Users Group- Maintained the existing infrastructure and developed new back-end features and APIs on the legacy LAMP (PHP) stack.
- Performed front-end work using HTML5, CSS, and JavaScript to improve user experience and fix legacy bugs.
- Wrote Python scripts to automate data processing and reporting from third-party APIs and internal database sources.
Technologies: Linux, Apache, MySQL, PHP, Python, JavaScript
Brian Todd
Natural Language Processing (NLP) Developer in New York, NY, United States
Member since May 20, 2019
Brian is an experienced Data Scientist and Machine Learning engineer with a track record of researching and deploying a wide range of models for natural language processing (NLP) tasks, computer vision algorithms, classical machine learning algorithms, Bayesian statistical models, time series analysis models, and large-scale data mining algorithms.
Brian is now available for hire
Portfolio
- Parabolica LabsPython, Cython, SQL, NumPy, Pandas, Scikit-learn, SciPy, AWS, Git, Hadoop...
- Twosense, Inc.Python, Cython, SQL, NumPy, Pandas, Scikit-learn, SciPy, AWS, Git
- SkedaddlePython, Flask, SQL, Snowflake, Pandas, NumPy, Scikit-learn, SciPy
Experience
- Python 8 years
- Scikit-learn 6 years
- Pandas 6 years
- NumPy 6 years
- SQL 6 years
- Data Science 6 years
- Computer Vision 4 years
- Natural Language Processing (NLP) 4 years
Availability
Part-time
Preferred Environment
Linux, Git, AWS, Sublime Text, Bash
The most amazing...
...project I've developed was a real-time deep learning pipeline that detected and classified user actions based on sensor data from a wrist-worn device.
Employment
Experience
- ABayesian (Development)https://github.com/brian-todd/ABayesian
Python package for Bayesian hypothesis testing.
- CyDTW (Development)https://github.com/twosense/CyDTW
High performance DTW library written in Cython for Python 3.x.
- FX-Data-Loader (Development)https://github.com/brian-todd/fx-data-loader
This project serves as an extensible framework for processing and cleaning historical FX data from Dukascopy's data feed.
- CFBBot (Development)https://github.com/brian-todd/CFBBot
Chatbot built to answer questions about several college football teams. Built using Rasa, Spacy, and requests.
- Patented Activity Recognition Algorithm (Other amazing things)https://patents.google.com/patent/US10548513B2/
A variety of techniques are used automate the collection and classification of workout data gathered by a wearable physiological monitor. The classification process is staged in order to correctly and efficiently characterize a workout type. Initially, a generalized workout event is detected using motion and heart rate data. Then a location of the monitor on a user is determined. An artificial intelligence engine can then be conditionally applied (if a workout is occurring and a suitable device location is detected) to identify the type of workout. In addition to improved speed and accuracy, a workout detection process implemented in this manner can be realized with a sufficiently small computational footprint for deployment on a wearable physiological monitor.
US Patent US10548513
Skills
Languages
Python, SQL, C++, Bash, SnowflakeLibraries/APIs
NumPy, SciPy, Pandas, Scikit-learn, PyTorch, XGBoost, Matplotlib, SpaCy, NLTK, Dask, TensorFlow, OpenCV, D3.jsTools
Git, Seaborn, Rasa.ai, Gensim, Apache ImpalaParadigms
Data Science, Unit Testing, MapReduceStorage
PostgreSQL, MySQL, Redshift, SQLite, Apache HiveOther
Machine Learning, Computer Vision, Natural Language Processing (NLP), A/B Testing, Mathematics, Statistics, Deep Learning, Data Mining, Cython, Google BigQueryPlatforms
Linux, Amazon Web Services (AWS), Unix, Google Cloud Platform (GCP), CUDAFrameworks
Flask, Hadoop, Presto DB, Spark
Education
- Bachelor of Science degree in Mathematics2009 - 2013University of Michigan - Ann Arbor, Michigan, USA
