Senior Data Scientist
IBM Corporation
مجموع سنوات الخبرة :10 years, 7 أشهر
Project Highlights:
• Knowledge Management Applications @Shell:
-- Shell company manages 7 applications for Knowledge Management (KM) to share knowledge, learnings, find experts across different domains, interact and collaborate with people in the organization
-- Responsible for data engineering and database management of all Knowledge management applications using Microsoft SQL server
-- Designed and built Power BI dashboards that captures various transactions, activities and usage statistics of each KM application to track the efficiency and productivity of the applications
-- Project management through Scrum-Agile methodology using Azure DevOps tool
• Chat with PDF files using LLM Chat Model @BP:
-- BP collaborates with various organizations across multiple business domains such as IT, finance, etc., Statement of work (SOW - pdf files) is created for each project explaining the scope of work
-- Lead a team in building an LLM chatbot for Business team using Retrieval Augmented Generation (RAG) through Langchain and Llama 2 LLM model and Streamlit application for extracting information from the SOW pdfs
• Electricity Consumption & Customer Behaviour Regression Model @AusNet:
-- AusNet is an Australian energy services company that surveyed their customers on different attributes and behaviour to understand their effect on energy consumption
-- Multiple statistical tests applied on the attributes like correlation analysis, t-test, analysis of variance (anova) to explain the effect of each variable separately on energy consumption
-- Regression model created using Python to explain through regression coefficients the contribution of each of customer attributes/behaviour to identify the attributes that significantly impact the energy consumption
Project Highlights:
• Market Mix Modelling @Lenovo:
-- Lenovo spends on different digital marketing channels such as SEM, SEO, CSE, Paid Social, Display, Direct, Affiliates promotions, etc., to sell their laptops.
-- Lead a team that built a Market mix model (Log-Log Regression) to optimize the spends in different digital marketing channels with respect to the Revenue(Sales) using different spend variables of all marketing channels.
-- Marketing response curves constructed for each channel which shows the diminishing effect of each spend on revenue to understand the breakeven point for a channel and ROI at different spend levels
-- Simulator tool built using Excel that incorporates the model coefficients and optimizes the channels spends for different business cases and scenarios using the Excel Solver
• Forecasting model @Lenovo:
-- Built a forecasting model for Lenovo to forecast their weekly revenue, orders and key digital marketing KPI like web visits, etc., to monitor weekly sales target attainment for laptops
-- Used multi variable forecasting model library(Prophet) that is built on deep learning neural network models for incorporating multiple variable effects for accurate forecasting
• App Migration Strategy building @Lenovo:
-- Built a migration strategy for Lenovo for migrating ~5000+ apps / reports / dashboards that are built in Tableau and Qlik Sense platforms to migrate it into Power BI
-- Extracted meta data of all the dashboards using a third-party tool that outputs the data on different dashboard attributes like measures, dimensions, parameter, calculated fields, filters, data source information, etc.,
-- Designed a data model to store, process and aggregate all the extracted meta data of the dashboards
-- Built a Power BI dashboard that visualizes the complete migration strategy along with the complete information on metadata and usability of all the apps and prioritization criteria for migration
Project Highlights:
• Resolution Recommendation @AT&T:
-- Created an NLP based Resolution Recommendation system for AT&T that predicts the resolution for network alarm tickets/cases to reduce the resolution time and cost associated with resolving the network issues
-- Vectorization techniques like Count Vectorizer, TFIDF Vectorizer and Word embeddings techniques like Word2vec applied for text pre-processing on the historically solved tickets using the comments (problem description) logged in the ticket
-- Cosine Similarity used to obtain text similarities between the new ticket and historical tickets to recommend multiple possible resolutions based on top 3-5 similar tickets
• Customer Churn Model @AT&T:
-- Developed a customer churn prediction model to identify the telecom customers who are likely to churn
-- Extensive EDA performed for feature selection of variables to be included in the model that likely differentiate the customer behaviour
-- Model deployed through Flask API into AWS
• Concentrix is a contact center company which provides customer care support and services to various clients across globe though calls, emails and chats
• As Senior Data Analyst of the company, the responsibilities involved
-- Storing and processing data in SQL servers for multiple regions like China, Europe, ANZ, etc.,
-- Designed and built Tableau dashboards that capture key contact center KPIs like utilization, occupancy, customer satisfaction%, call handling time, agent unplanned absenteeism, etc.,
-- Prepared weekly/monthly governance decks as part of data reporting to
sites leads with valuable insights to identify key focus areas to improve the business and process to maximize revenue, quality and efficiency
-- Sentimental analysis through NLP techniques on customer feedback to capture the satisfaction and dissatisfaction drivers to enhance customer experience
-- Forecasting call volumes for capacity planning through forecasting model such as ARMA, ARIMA, etc.,
Project Highlights:
• Customer segmentation / classification models for a telecom company using kMeans clustering algorithm, Customer Lifetime Value, Customer Churn model
• Developed Tableau dashboards for Sales and Consumer analytics that provide insights to improve sales, revenue and customer satisfaction
Bachelor of Technology (B.Tech) - Mechanical Engineering