Anshuman Pradhan Email and Phone Number

• From Scratch, developed and rolled out a pioneering GPT35Turbo based “Agent Email Reply Recommendation Engine”, tailored to generate recommendations based on historical replies to similar cases, requiring minimal modifications. • Successfully rolled out to customer service agents across 144 countries spanning 15 languages supported by the company within Maersk's CRM system, the solution is poised to save 300,000 man-hours annually, significantly enhancing operational efficiency and customer satisfaction by reducing Turnaround Time (TAT).• Designed and deployed a real-time CRM case triaging model for Salesforce, exceeding the performance of the legacy Einstein model with an accuracy increase from 52% to 86% within three months, assisted by a batch retraining pipeline. Resulted in a reduction of $1M in Salesforce licenses and 73 Full Time Employees in the CRM workforce.• Demonstrated expertise in AI by filing various patents, addressing challenges including logo identification and deletion, attachment information extraction, real-time dynamic issue identification, and generating real-time email reply recommendations.

• Delivered impacting products like Contextual Search and Tamper Detection on scanned contracts using OCR and NLP algorithms.• The on time delivery of Contextual Search Engine was attributed with savings of $460M worth of revenue leak due to bad clauses.

1. Trade Oversight and Supervision:• Developed a Risk Index metric to quantify advisor behaviour in suggesting funds for customer trades.• Translated stakeholder requirements into a data science problem, resulting in a pattern checking algorithm.• Identified anomalous trading patterns, leading to immediate deployment of the algorithm.2. Article Recommendation Engine: • Implemented a Reinforcement Learning-Multi armed Bandit based algorithm for content recommendations.• Innovated a solution for the article cold-start problem, leading to a patent application.• Achieved a 40% increase in account openings through A/B testing of the algorithm.3. ML based Representative Assignment :• Designed a Customer-Advisor ranking algorithm for instant representative allocation.• Reduced time complexity of advisor/customer search problem and deployed on AWS Sagemaker.• Automated advisor reassignment, saving significant man-hours and reducing attrition.4. Propensity Models for Cashback Offers :• Utilized data science to target HNI investments effectively.• Pilot predictions generated a 40B $ opportunity in new money from HNI in one year.

Robotics, Autonomous Vehicles.

- Played a key role in research and development efforts aimed at developing biometric authentication systems using ECG signals.- Applied signal processing techniques to deduce prominent features from ECG signals for biometric signature storage.- Utilized Support Vector Machines (SVMs) and clustering algorithms to effectively recognize individuals based on their ECG signals.

Led a team of 30 members from different branches at IIT Kharagpur to build an autonomous self-driving car, aiming to be the first of its kind in India.

- Extensively used Machine Learning algorithm to classify multiple classes of Arrhythmia with an accuracy of 98.3%- Used Support Vector Machine (SVM) and Decision Tree algorithms for training classifier- Authored a paper “Smart-phone based Real-time Single-lead ECG Arrhythmia Classification” in COMSNETS-2014, Bangalore