Insurance

1. Insurance cost prediction for home care appliances: Predicted the renewal insurance amount for homecare devices such as room heater and boilers, central heating, plumbing, etc. prediction was performed based on manufacturing years, energy unit consumption, previous insurance amount, maintenance history with cost and credibility of customers for different locations.

2. Use of Predictive Analytics for detecting propensity of fraudulence of a policy: Using both Policyholder’s data and Agent’s data to predict fraudulence. The project requires being innovative in creating features/variables and use of ensemble learning and back testing to reduce chances of overfit.

3. Smart customer management system: Managing customer communication via text-based channels (emails, SMS, chat, social media, etc.) is labor intensive and inefficient. Create a tool on taxonomy of current business model and customer query which assign to right query handler, thereby, Optimize the cost by reducing manual work.

4. Retention model for Insurance policies: This model was made for an auto major to predict the retention probability of customers using application and monitoring level data. The model was made using MSSQL server and R and deployed on shiny.

Manufacturing

1. Battery Failure Prediction: Using ML techniques to predict vehicles having battery issues such that a better targeted service campaign can be used for improving customer services.

2. Chemical quality prediction: The goal of this project was to predict the quality of chemicals based on the features and learn how each feature impacts value. The dataset was related to red variants and white variants chemical. Features consist of pH values, citric acid, chlorides, Sulfurdioxide, density etc captures by using various test. The output is median of at least 3 evaluations made by experts.

Retail

1. Churn model for Ecommerce: This project was deployed for both online platforms and offline stores using a probability based approach and a sensitivity of the customers to the different campaign variables under our control to maximize customer lifetime value and minimize churn yet keeping the campaign cost low.

2. Social Media Brand-Influencer Platform: The platform is built as a complete end to end system with a python – mongo (LEMP stack) with a R shiny front end. The application gets data from social media platforms using API and scraping and then uses advanced ML/NLP and AI algorithms to select the best influencer match for brands and products as well as vice versa.

3. Marketing Mix Modelling: An E-commerce project that involves finding customer's path to product purchase on amazon.com. Further providing an effectiveness analysis on all platforms where the client advertises their products.

4. Product Review Ranking: The project was to rank the helpfulness of user written product reviews by extracting features like degree of objectivity, subjectivity and reviewer profile from the review text.

Banking

1. Credit Risk Models for Wholesale and retail portfolios: Developed PD, LGD, EAD models and capital estimation modules completely in SAS and MS Office for large MNC banks. This included validation models as well. The projects included complete work from data sourcing from different databases, ETL, modelling validation and developing a connected visual app in the front end.

2. Collections strategy: Predicting customers with the highest probability to pay after being delinquent. Using a combination of demographic, bureau and social media variables to build a predictive model.

3. Income estimation model: Estimating the income segments for loan applicants based on application information and asset pricing. Since accuracy rates of point estimation models were low, a segmentation model was built to reduce physical verification costs.

HR Analytics

1. Predict Employee Turnover: Project was built by using Python Scikit-Learn library. Machine learning models such as logistic regression, support vector machines, Random forest etc were used to develop models for employee turnover based on attributes such as hours spent at work place, no of leaves availed, department, relative level of salary ,appraisals etc.

2. Performance Prediction:This model was developed to identify traits, behaviors or team activities that explain good or bad performance in individual or team settings. Usefulness of this tool was to figure out issue that impacts performance even before any deterioration in performance occurs.

3. Resume screening: This tool uses average word embeddings model to search relevant resume through the database created through internal db, job portals api access and scraping. Domain-trained word embeddings with pre- trained embeddings for resume.

OPTIMIZATION TECHNIQUES

1. An optimization model was developed for maximizing tyre dispatches& minimizing the cost of buses and maximizing Depot compliance, developed using CBC solver with having a complex system of constraints and rules.

2. Built an optimization model for reducing waiting hours by passengers for trains by reclassifying the stops and the stop time based on footfall at each station.

3. Built an optimization model to calculate the minimum distance calculation based on geolocation codes for a network to be active in a meetup group.

OTHERS

1. Support vector machine was used to predict & categorize customer complaint tickets into different modules to increase operational efficiency.

2. Image recognition techniques were used to classify images from an online property portal and derive dynamic pricing structures.

3. Using Mobile sensor data, derive driving behaviors such as over-speeding, rash driving and roads profiling to help insurers get better Resources into their portfolio and devise both preventive and prescriptive policies.

9

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.