Mohit Bansal

Collaborated with engineering and product teams to launch a flagship seed product recommendation model in an iOS app for customers across 13 US states. Led a team of 2 contractors and worked with engineering teams to successfully plan and execute the deployment and validation of a prediction model onto production systems. Increased known rate by 5% by adding a cold start algorithm to the recommendation model to find proxy products in order to make recommendations against unknown and new competitor products. Saved the organization ~$250,000 by analysing the impact of a data gathering program on the final training dataset using statistical tests and recommending its discontinuation. Reduced training time by 90% by implementing data-parallel training for a Bayesian Probabilistic Prediction model with Spark leading to accelerated development and validation processes. Increased data acquisition by 10% YoY by developing a machine learning model to predict the volume of agronomic data uploaded by users which aided in the creation of a personalized marketing campaign.

Delivered insights for over 100 seed products by performing EDA of on-farm data leading to the creation of a recurring report delivered to several teams for better data-driven decision making. Published 20+ dashboards in Looker to analyze diverse metrics, and discover correlations enabling improved decision-making for the product, commercial and other teams across the organization. Defined and computed executive metrics for the company leadership and quantified the impact of poor data quality that could be addressed to meet year-end as well as strategic goals of the company. Established standard operating procedures for dashboard publication within the team and acted as an on-boarding buddy for new hires and trained them on team and company technology stack.

Collaborated with cross-functional Supply Chain and Warehousing teams to automate the reporting of forecasting process using SQL and Tableau and eliminated manual effort by 75%. Increased availability by 20% at 9 warehouses by calculating utilization rate and fixing data quality issues.