Nirmaya Narayanan

Involved in several structured and ad-hoc analytics and planning projects.

• Contributed to a project that recommended a Booking quota at Customer level using historical booking patterns. Contribution was grouping customers into several clusters based on customer characteristics like previous orders, historical revenue generated for the company, geographical location, business type, etc., using Clustering algorithms. K-means clustering and t Cluster definition parameters were used to achieve ideal clusters.
• Worked on a Customer lifecycle Engagement Project to identify key risk indicators of major down-sell or no renewal of opportunities in pipeline. Used existing project outputs to bucket opportunities into low, medium and high risk for non-renewal and automated the previously manual process with improved accuracy of 80+%
• Worked on a Text Analytics project to derive insights from feedback data. The company collected associate's response on 'Changes in Travel Policies' globally after the advent of COVID-19 and textual feedback was recorded. Derived insights using sentiment analysis, word clouds and other data drilling techniques including major concerns to be taken care of, region and department level issues, FAQs, which helped Management to create Focus groups and address the issues.
• Onboarded and extended support to run and improvise a project that created compliance scores for customers in order to facilitate easier inspection of compliance and retrieve revenue from non compliant customers. This saved several weeks of man hours every year for the company.

• Worked on a project to use textual updates from sales representatives to build a predictive model on deal closing. Created multiple features from text data including word count, frequency of updates, word clouds, sentiment analysis, etc., and built a Random Forest Model with 75% accuracy to predict the direction of an opportunity in its early stages.

• Was part of creating a Computer Vision model for a Car Insurance company using Image annotation and Machine Learning. The project aimed to build a model that could identify car part damages and assess if the damage qualifies for an insurance approval. This was achieved by training a Model with previously annotated image data of damaged and intact car parts from different angles. Managed to Annotate Data for 20 types of Car models with 60+ parts