Mike Stefkovich Email & Phone Number

Richmond, Virginia, United States

District Of Columbia, United States

Served as a consultant providing strategic advice, product design recommendations, data analysis, code implementation, experimentation, and ongoing support. Produced a python application which calculated the yield of competing reservation requests used in sales and revenue operations.

New York, New York, United States

As the Live Video Experience D.S., I influence strategic investments, identify opportunities, and measure team success. Some highlights include reducing daily notifications sent by over 1 million, discovering causal links to topline metrics via feature experimentation, and influencing the reallocation of resources to comment improvements & broadcaster monetization features.

New York, United States

As a data scientist at Centene, I was tasked with working on the self service chat agent. This included training and evaluating various iterations of our Rasa NLU model. Of particular note is the company and domain specific word vector model I trained using AWS. The end result was $500k+ in annual call center savings.

Tysons

As a data scientist at Tegna, I was tasked with creating multiple end to end data driven products to forecast time dependent buyer behavior. Using tools like apache-airflow, I scheduled and automated time-series workflows I built within Python using Facebook Prophet. The resulting product replaced a legacy SAS process, saving $300,000 in annual licensing while decreasing forecast error and reducing compute time. I implemented the solution after determining minimum AWS system requirements to… Show more As a data scientist at Tegna, I was tasked with creating multiple end to end data driven products to forecast time dependent buyer behavior. Using tools like apache-airflow, I scheduled and automated time-series workflows I built within Python using Facebook Prophet. The resulting product replaced a legacy SAS process, saving $300,000 in annual licensing while decreasing forecast error and reducing compute time. I implemented the solution after determining minimum AWS system requirements to achieve data refresh cadences of our parallel processes. I encouraged 40+ national television stations to adopt and trust a weekly issued rate card model by carefully incorporating rules based feedback. Show less

Tysons, Va

I performed multiple forecasting tasks while working within the revenue management arm of the Data & Analytics Team ("DNA") team at Hilton Hotels headquarters, Using Python I forecasted booking channel mix while achieving an average monthly MAPE below 1.5%. I independently conducted the revenue impact calculations of the 2018 Chicago hotel strike, AAA discount product, and the 2017/2018 international weather events. Lastly, I created dynamic Tableau visualizations and automated legacy reporting… Show more I performed multiple forecasting tasks while working within the revenue management arm of the Data & Analytics Team ("DNA") team at Hilton Hotels headquarters, Using Python I forecasted booking channel mix while achieving an average monthly MAPE below 1.5%. I independently conducted the revenue impact calculations of the 2018 Chicago hotel strike, AAA discount product, and the 2017/2018 international weather events. Lastly, I created dynamic Tableau visualizations and automated legacy reporting with R and Excel, freeing up 125+ man-hours annually. Show less

Fairfax, Va

Worked on an internal research and design contract to predict adversarial network traffic density with neural networks. Achieved classification accuracy improvement by 38.6%. In addition to forecasting, I automated OS and software installs using PXE boot and Ansible.

Fairfax, Va

Experimented with and developed a flight path linear regression algorithm within a virtual machine cloud computing environment utilizing a hadoop data warehouse.

Keuka Park, Ny

Implemented several Python machine learning models to aid in revenue focused decision making. Keystone projects included prospective client evaluation via a K-Nearest-Neighbors algorithm and a random forest fit dynamic pricing model. Supporting work included Selenium web-scraping, data pipe-lining, and task automation.

Rochester, New York Area