Keerthana KS

Leading the NLP Domain of the AICOE.

Managed, mentored and empowered datascience teams of size 2-5

• Implemented an NLP based Intelligent daily newsletter for the organization using Document similarity based recommendation engine and Fast Api UI.
• Autoclassified complaints and inquires and provided unknown insights from unstructured data using various Topic Modellling algorithms.
• Lead a team to have POC on Automatic speech recognition for UBP chatbot with speaker identification.
• Working on a research paper to find the best NLP technique based on topic noise variation.
• Collaborating with Queens University, Canada for a XAI paper on Identify the explainability preferences for lending officers.
• Worked on Mules and Phishing fraud detection using XgBoost
• Lead and conducted L&D sessions to upskill employees and interns.
• Mentored 4-5 interns from SMU and NUS singapore.

• Collaborated closely with client to derive the CLTV (Customer lifetime Value) for a period of 5 years for all markets across globe for Philips.
• Performed Logistic Regression to predict the likely hood of repeat purchase for owners and prospects for three years data and fitted the model on next 2years data.
• Used Collaborative Filtering as a Recommendation engine to predict next product purchase based on past Purchase(AG’s)
• Calculated Volume of purchase/ product units from the past purchase behavior.
• Used Clustering to do a Look Alike analysis with owners to predict the Value of Purchase for the prospects.
• Calculated the CRM and Advertising costs for all Markets to derive at the CLTV by deducting these cost from the overall Revenue obtained from the above methods.

Market basket analysis For Targeting and generate Recommendations:

• Identified products and accessory groups at the AG level, which customers tend to buy or browse together to send targeting marketing notifications to customers based on their past purchases.
• Increased cross sell on platform by communicating relevant products to particular customers based on customer Last purchase / browse history.

RFM (Recency, Frequency and Monetary analysis):

• Identified customers who are more likely to respond to promotions by segmenting them into various categories.
• Examined the RFM scores to know how recently a customer has purchased, how often they purchase and how much they spend.
• Built an end system for campaign team to receive the RFM scores for each segment of customers

Application Development and Implementation Management