I teach Foundations of Data Analytics, a course centering on Python, probability, and a short survey of industry techniques for clustering, NLP, and time series analysis

- Created system from the ground up to support rapid model deployment and iteration, science to production data streams, and accurate analytical BI and go-to-market reporting- Built and standardized the system for the DS, ML, and Analytics teams to start, test, and deploy production ML and statistical models using Poetry for package/venv management and templatized projects in Github. Enabled the team to generate IP in 9 greenfield spaces including recommendation, matching, ranking, regression, classification, retention/churn modeling, and causal impact.- Led cross-functional data infrastructure meetings to plan, build, and coordinate Data/MLE/Eng design and construction of pipelines and infrastructure strategy for robust Science, Analytics, and production tools.- Constructed a custom internal Python library with custom database connection classes, virtual environment management via Poetry, app prototyping and internal deployment via Streamlit and Docker- Worked directly with VP of Data Science on DS roadmapping, IP strategy, and team resource allocation- Mentored 3 data scientists, focused on algorithm development, machine learning model deployment, code review and best practices, and solving complex, open-ended problems

- Managed a team of 6 data scientists across levels focused on modeling store inventories, overhauling infrastructure, and building solutions to support the human-in-the-loop Operations team.- Defined Q2 and Q3 roadmaps, led the team in sprint planning, owned quarterly objectives and key results, and reported progress directly to CTO and Head of Product.- Provided technical and scientific coaching to launch a controlled inventory experiment. Achieved $1.3m in annual savings for the company.- Oversaw deployment of a novel simulation model to 100k SKU-locations that reduced inventory discrepancies.

- Led a scrum team consisting of 3 data scientists and 3 engineers over 4 months to fully refactor production forecasting code. - Developed and reviewed technical specs, conducted intrateam paired sessions, expanded end-to-end testing, and rigorously documented the new framework.- Created a robust, scalable machine learning ecosystem in Flask deployed via Azure CI/CD.- Designed and built a custom Python class to unify demand model training and deployment under one framework in collaboration with engineering.- Introduced rigorous Pandas dataframe type-checking throughout Python forecasting code using Pandera.- Built a Python model to forecast grocery deliveries using historical waste and delivery data. - Constructed and deployed a greenfield solution to improve holiday orders by predicting demand deviations.- Led the team in adopting and investing in unit testing, growing the test base from 0 to 400+ unit tests, eradicating 90%+ of fires in 6 months. Ran a weekly training workshop for other data scientists.

- Lead data science investigations to support and improve Outlook universal- Work with PMs and engineers to diagnose and solve various problems with user telemetry and app usage- Use Jupyter notebooks to analyze and report trends in user engagement and retention- Write and execute Scope queries to aggregate various large data streams for analysis- Advocate for more data-driven approaches to underlie vital business decisions

- Created a web application (http://baseballcontracter.science/) that predicts which baseball free agents will receive contracts and the annual value of those contracts based on player performance and league factors- Scraped and extracted performance statistics for 2000+ free agents using Python (requests, BeautifulSoup)- Built a PostgreSQL database with 10 tables of player statistics, team performance, and contract information- Trained logistic regression models for predicting which players would receive contracts and linear regression models for predicting contract values for pitchers and position players in Python (Scikit-learn)

- Built a suite of software tools and resources to help better model interactions between genes and how they give rise to different phenotypes such as Alzheimer’s disease- Created an R package to assemble over 400 million DNA footprints for training logistic regression and gradient-boosted classification (xgboost) models that predict gene expression- Developed a Bioconductor package that provides 9 different methods for performing feature selection on gene regulatory networks to hypothesize which genes are most likely to interact- Created 108 PostgreSQL databases of DNA footprints covering 27 different body tissues using parallelized functions in R (BiocParallel) to enable construction of different tissue-specific gene networks- Enhanced a Bioconductor package that collects motifs from various databases and other sources, by parsing and adding 5 new databases and 1 additional output format

- Created and taught a recurring 7-week course to 45 cancer biologists covering intermediate concepts (dplyr, loops, functions) in R to help them better conduct data analysis- Developed additional materials to better engage and test student understanding with real-world data problems (https://marichards.github.io/FH_intermediate_R/)

- Built a MySQL database of over 750 manually-extracted microorganism growth conditions to better understand how to grow difficult-to-culture organisms (link below)- Developed a model in Matlab to predict ways of converting methane, a greenhouse gas, into liquid fuels- Directly supervised and mentored 10 undergraduate students

- Developed a procedure to clean soiled hot melt extruder parts- Designed and executed dip test experiments with different cleaning solutions and sonicator baths.