ETL & Reporting tools:
Situation : Identify attacks in IoT devices
Task : Build an anomaly detection model using machine learning or Deep learning models
Action : Stimulated IoT behavior using the sample data and performed feature engineering to increase the performance of the model. Built auto-encoder, robust auto-encoder, isolation forest in unsupervised learning and random forest, gradient boosting in supervised learning models.
Result : The model predicted IoT network attack with 92 % accuracy after feature engineering using the combination of both supervised and unsupervised learning models
Situation : To detect the utility poles on the streets for deploying 5G radio devices
Task : Built an image recognition model to detect utility poles from street view images
Action : Downloaded 10 thousand street view images from bing and created a training dataset. Built CNN model using Resnet50 architecture and weights to detect utility and streetlights in the images. Performed active and transverse learning to train and test the model
Result : The model was able to predict the utility pole with 90 % recall after tuning the hyper parameters.
Situation : To identify the most valued customers
Task : Identify a methodology to create a metric to calculate the customer's value
Action : Created a customer lifetime value metric (LTV) by assimilating marketing expenses, mail campaigns and customer purchases. Created product LTV to identify the important products and manage the inventory efficiently. Perform A/B testing to understand the impact of promotions
Result : Increased the coupon code utilization by 15% and increased the sales by 5% in the Q2'2018
Situation : Identify the areas(/region/county/state) to improve sales of list of products
Task : Build the ML model to predict the sales and also to identify the important feature
Action : Created a regression and logistic regression models to improve the sales and churn. Perform recursive feature elimination with cross validation (rfecv) to identify important features and to improve the model performance. The model is compared with RMF (Recency Frequency Monetary Value)
Result : Model was able to predict the sales with 89% of accuracy and identified the areas to improve the sales.
Situation : To reduce the attrition of the employees
Task : Built a ML model to identify the areas to improve the attrition
Action : Collect the data from multiple sources, prepared required data for analysis. Created multiple ML and DL models to identify the areas affecting the churn. Created the benchmarks for data collection and reports from the employee feedback.
Result : Able to reduce the attrition by 5%, created and automated the report for regular validation
Zero Based Budgeting (ZBB)
Situation: To create a budget using ZBB concept
Task: Built ML model to forecast the budget and reduce the expenditure of the company, monitor the expenditure on a weekly and monthly basis
Action: Create the dashboard using Tableau to create an annual budget and periodic forecast/outlook and Generating insights. Built theARIMA model to create a time series forecast to predict the budget. Scrutiny of General Ledger on a monthly basis and do a financial analysis o
Result : Reduced the company's expenditure by 1 million dollars and successfully able to monitor the budget
Data Management in the life Sciences Industry
Situation: Create an incentive plan health care, representatives
Task: Built data pipeline to collect data from multiple resources and create a data management model to calculate incentive
Action: Develop a SAS data pipeline to integrate data from multiple resources and check for data compatibility. Develop dashboards to integrate the data pipeline to show the updates of the incentives to sales representatives.
Result: It helped the health care representatives to easily understand the incentive split and challenge them to improve more sales of the drugs
TXU Big Data
Situation: Reduce the customer attrition
Task: Built ML model to understand the reasons for attrition of customers
Action: Developed Logistic regression, decision tree models, preformed feature importance to reduce the features to reduce the processing time and to improve the performance of the model. Understanding the oil and gas industry to understand the features.
Result : Successfully was able to identify the areas affecting the churn in the electricity industry.
Resumes, and other information uploaded or provided by the user, are considered User Content governed by our Terms & Conditions. As such, it is not owned by us, and it is the user who retains ownership over such content.
Companies Worked For:
Job Titles Held: