Experience

• 

  Infosys ltd

  Jul 2011 - Sept 2013

  System engineer

  Played Key Role in planning and setting up org wide Management server components for automated software deployment and maintenance, automated report generation of the health of IT services, active directory setup, and SQL DB configuration and setup.Tools Utilized: System Center Configuration Manager, System Center Operations Manager, Active Directory, MS SQL Server, Altiris Deployment Solution.

• 

  Nuance transcription services india private limited

  Apr 2016 - May 2016

  Data analyst intern

  Designed a Linear Regression Model which measured effect of Speech lines transcribed, hours spent on transcription by Medical Language Specialists, Quality Check L1 Escalation and Quality Assurance L2 Escalation on Fixed Pay Loss. The model was used to study and understand the interaction effect between these variables and determine the most impactful loss.Tools Used: Excel, VBA

• 

  Latentview analytics

  Jun 2017 - Sept 2020

  Senior data analyst

  • Conceived the design of A/B testing strategies using Bayesian hypothesis testing that improved digital rewards program efforts and boosted true conversion rates by 47%.• Initiated a comprehensive metrics framework of 100 KPIs and utilized density segmentation algorithms to generate customer segmentations; adopted into all future campaign designs.• Refined Digital Loyalty Program ROI impact measurement to adopt Bayesian Structural Time Series; isolates true impact of loyalty program to 13%-18% on organic search revenue; facilitates decision-making for senior executives in yearly financial reporting.• Redesigned and implemented major strategy change to A/B testing cohort selection methodology; utilizes 25+ behavioral and lifetime KPIs to segment cohorts; reduced campaign costs by 35%.• Proposed Uplift Propensity Score modelling to isolate segment jumps while measuring incremental lift; PoC accepted by senior executives. Show less

• 

  Bridgei2i analytics solutions

  Oct 2020 - Feb 2022

  Program manager

  • Conceptualized and spearheaded a cross-functional team for an end-to-end AutoML platform development reducing ML workloads by 35x; scalable up to 10 models per user; increased ROI by $60M+ YoY from a single business unit.• Conceptualized and implemented a Causal Inference Root Cause Identifier engine to measure after-campaign failure reasons and improving future campaign targeting; reduced targeting costs by 40%; improved campaign conversion rate by 65%.• Supervised a 3-man team of data scientists that designed and implemented a recommendation engine for 20+ insurance products based on 50+ metrics; improved up-sell by 35% and cross-sell by 18%. • Conceptualized a cloud native plug and play claims fraud management platform; clients can seamlessly plug their data and utilize existing trained models; internal teams can plug trained models as new services; spearheaded 4-man team to build some models in the architecture.• Led the Insurance Operational Excellence team and coordinated with the HR team for resource planning, utilization and monitoring for all projects under the Insurance Business Unit.• Mentored junior members and coached new joiners; taught workshops on ML technology and theories; conducted 50+ interviews. Show less

• 

  Verizon

  Feb 2022 - now

  Senior data scientist

  • Reduced fraud loss by $30M+ and false positives by 25%; implemented and maintained 10+ Deep Learning models using Pandas, Scikit-Learn and Keras.• Designed and automated a Loss Prioritization Report using Pandas and Seaborn to monitor fraud loss based on 30+ internal segments for Retail customers; adopted by 10+ senior executives as bi-weekly reports.• Designed and developed a Skill Optimizer and Project Resource Allocation Engine using Julia, Julia Data Frames and JuMP for processing up to 10M employee skill records and project level constraints; cutting down planning time by 140x. Recommends upskill, cross-skill and optimal resource per project.• Designed and developed an AutoML Pipeline with Bayesian Hyperparameter tuning which generates an optimal model with incremental fraud capture rate of 20%+ under 50 trials with minimal size artifact generation; all in under 4 hours.• Introduced code base standardization and overhauled old infrastructure and dev environment management; reviewed team’s designs, algorithms and code in an effective manner increasing job efficiency by 55%.• Mentored 10+ junior members and coached new joiners; conducted 20+ interviews. Show less