Aasheet Kumar Email & Phone Number

•Lead statistician of experimentation at Schwab in statistical model testing, analysis, design, and result interpretation.• Supervised experimentation onboarding with +/- 25 teams across various departments lending knowledge on statistical test modeling, power calculations, errors, and A/B test result utilization.• Transformed over 90% of unstructured experimentation data from Optimizely, JIRA, Confluence, SharePoint, Adobe Analytics, and Observe Point JSON API calls into tables using Python (Pandas, Beautiful Soup, PySpark, NumPy, regex).• Designed and implemented Bayesian A/B testing models utilizing Markov Chain Monte Carlo (MCMC) for testing teams with limited traffic volume using PyMC3.• Built a regularized regression machine learning model to forecast predictions on test build timelines as the first step in an initial experimentation ideation phase.• Generated reports for A/B test launch and maintained over 30+ dashboards on ongoing KPI results that are viewed weekly across Schwab (50% managers/directors, 20% individual contributors, 30% testing teams).• Managed Hadoop Tableau report connectors• Wrote and produced +20 data-scrapping/collecting scripts with Python and API JSON requests.• Trained seasonal ARIMA time series machine learning models on historic data to predict outcomes for an annual impressions budget with a +/-10% margin of error, reducing impression cost.• Built and Established ROI calculations from A/B test results via Pandas and NLTK packages• Calculated metrics between 2020 – 2023, wherein 174 experimentation tests were conducted with a sample size of 16m (4.5% of schwab.com) users resulting in:o conversions increased – 5.78% increasing conversion rate with a $8.2m 5yr NPV Gross Margin.o conversions saved – 2.79% loss averted with a $2.8m 5yr NPV Gross Margin Saved.o cost savings – 2.8k dev hours avoided estimating at $150/hr• Created custom power calculations for sample size and statistical significance A/B testing calculators

Houston, Tx

• Applied natural language processing on superintendent’s notes to analyze wharf crane downtime on a component level. Identified spreader bars contributes to more than 36% of downtimes.• Performed survival analysis on spreader bar usage to assess probability of survival by hours of usage. Identified certain spreader bars that have low survival rates and attributed it to their maintenance history• Developed alternate key performance index for wharf crane operations based on downtime by adding data from superintendents’ notes using sentiment and time series analysis• Analyzed freight movement of containers through Texas by trucks to identify high density locations and potential growth to make a decision on the optimal location of of new inland containers port saving up to 13 million miles of truck movement annually using Google geolocation, Directions API, shapely and other geospatial libraries.• Applied Convolutional Neural Nets on operational camera footage to provide information on improper cargo and wharf usage as well as to classify cargo type and location.• Analyzed light duty vehicle fleet usage statistics to provide statistical inference to lower vehicle fuel usage and inform future fleet recommendations.• Performed time series analysis and hierarchical clustering on the Port’s electricity usage data to visually represent demand patterns to identify potential energy efficiency improvements and assed the opportunity to procure renewable electricity.• Created dashboards for projects to provide better understanding using d3.js and tableau• Conducted ad hoc analytics on various requests involving deep exploratory analysis for operation, maintenance, infrastructure and financial departments.

United States

• Performed Survival and ANOVA analysis on patients with epithelioid mesothelioma to better predict survival for patients using tumor infiltrating lymphocytes.• Created Dashboard for Amazon seller’s ad effectiveness based on the target audience and customer nuances to improve advertising costs per sale.• Performed Man-Whitney U test on proposed algorithm before and after implementing constraint optimization to signify improvements.• Analyzed survey results and Statistical data using natural language processing in tableau dashboard for college and high school athletics departments.• Analyzed machine recalls data and attributed them to specific part failures and found methods to avoid such defects.

Houston, Texas Area

• Data analysis and visualization to Improve the utilization calculations for all the vessels coming into the port of Houston to accurately predict past and future utilization, seeing a difference as high as 20% between old and new calculations• Data visualization by creating a heatmap to depict the most used parts of the terminal using vessel visit length and footmarks to improve material handling and resource utilization.