Data Scientist Lead specializing in Supervised Learning, but primarily focused on Forecasting. Rick built a scalable state of the art Machine Learning model to replace hundreds of custom models which led to dramatic improvement in site level forecast accuracy (where +30% error was common to <15% error). Rick's predictions power all kinds of business critical initiatives ranging from labor staffing, zone skipping to automated Inventory Placement. He carefully brings state of the art technology into a constantly changing environment where new fulfillment centers are opened on a weekly basis, enterprise customers lauch new products and move to new locations on a weekly basis and models must carefully extract a complicated signal from a noisy data stream. This kind of robust performance requires so much more than machine learning expertise: business accumen, reliable engineering and a practical business-first mindset are essential. Besides machine learning Rick has authored numerous advanced anomaly detection methods, built custom pipelines to support promotions, inventory movement, customer churn and onboarding along with stock-out adjustments. He is capable of doing the work himself as well as turning clever junior engineers and scientists into productive builders through meticulous mentorship and patient teaching.

As a Senior Data Scientist specializing in real time pricing I was the Expert maintainer of various machine learning and statistical models that power the live pricing service and multiple business critical services (market-based carrier rep compensation, auto tender acceptance, auto spot bidding, book it now, etc.). Rick led development for automated tender acceptance, as well as a new pricing service that could estimate MarketCost achieving the same error as the best ensemle of many models while only requiring 5 input features common to all shipments. This kind of practical improvement is the difference between an unusable prediction and a reliable enterprise-grade service. Rick was able to make important breakthroughs in pricing acuracy and interpretability within complex environment where hours of downtime could result in six figure losses. Rick created a scalabe REST services on AKS implementing batch and shapely endpoints for extreme speed and interpretability when needed. He interviewed, hired, and mentored all applicants and new team members, helping to grow a team of 2 to over 10.

As a Data Scientist Rick deployed numerous statistics and ML-based pricing models using ensemble/boosting techniques, LightGBM, XGBoost, Catboost, Random Forest Quantile Regression, KNN-regression, etc. Rick helped tune a two-way lightgbm based recommendation model and demonstrate lift, implemented a shapely endpoint to give human readable explanations for model recommendations that give confidence and clarity, helping end users to get actionable insights that faciliate meaningful conversation points and warm leads.

As a Junior Data Scientist on a newly formed team, Rick built and maintained custom enterprise account forecasts with R and SPSS (ARIMA, ETS, etc.), he automated automated freight matching algorithms with R and Python and fostered meaninful communication between DBAs, Data Scientists and Software Engineers to create multiple iterations of collaborative filtering, statistical likelihood and machine learning based based recommendation engines. Developed and led AB test to determine the best algorithm to power live MyMatches application with a few hundred users.

Rick led the development of predictive model for 7 different debt portfolios(Healthcare, Utilities, etc.), he built numerous classification and regression models to segment debtors based on likelihood to pay and total expected value in an environment where <1% of observations had positive expected value. He deployed a scoring platform for incoming placements and performed statistical analysis to optimize revenue collection strategy.