Feb 2023 – Sep 2025 · 2 yrs 7 mos · Bengaluru, Karnataka, India · Hybrid

• Building AI & ML solution towards digital healhtcare

May 2020 – Feb 2023 · 2 yrs 9 mos · Bengaluru, Karnataka

Oct 2017 – May 2020 · 2 yrs 7 mos · Bengaluru, Karnataka, India

• Willingness to Pay Model
• Built a model which will predict the amount an owner can pay for package owner wants to buy.
• Increased conversion rate of process by 120% and average order size by 10%.
• Model is implemented using Kafka & Spark (real time streaming).
• Lead Engine to Sellers (Project Valuation- 10m USD)
• Built a model which will create a requirement & availability score. These leads will be passed to sellers such that they can contact buyers directly (basis on their property requirement).
• Using the scoring- Availability module accuracy increased from 50% to 75% and requirement module accuracy increased from 18% to 35%
• Home loan Conversion model using property data (Project Valuation- 5m USD)
• Built a data science model which has increased conversion rate from 0.5% to 3% for home loan using only internal property database.
• These leads are passed to banks, DSA which can enhance company revenue by 10%-15%
• Increase average order size & conversion rate for B2C & B2B process
• Built Willingness to pay and prioritization model using machine learning algorithm to increase average order size & conversion rate respectively
• Current conversion rate has been increased to 16% from 2% and average order size has been increased to 7K from 4.5K
• Customer call pickup rate has increased to 72% from 64%

Jan 2016 – Oct 2017 · 1 yr 9 mos · Bengaluru Area, India

• Loan Singh
• Currently building a alternate credit model using telco-data
• Built alternate data credit model using education, company, social, demographics, and financial factors of a borrower for professional certifications, personal loan and telecom payment bank
• Built an unsupervised learning model to score 32,000 colleges and 760,000 companies in the Indian market
• 0% default rate post usage of model (current disbursal rate>1 crore per month)
• This solution was built using unsupervised learning models and is currently being validated against industry benchmarks in real time market
• PointsWala
• Built recommendation engine using demographics, social, consumer preferences such as brand, price, discounts, color and material, collaborative filtering. This predicts best fashion offer for a particular user using aggregated offers from Flipkart, Amazon, Myntra, Jabong and many more.
• This solution was built using unsupervised learning models, Neo4j, graph database and theoretical concept of deep learning

Sep 2014 – Dec 2015 · 1 yr 3 mos · Bengaluru Area, India

• Customer Targeting
• Developed analytical solution to increase the targeted response rate for two different insurance products. Initially the response rate was 2.8% and 0.6% respectively which increased to 5% and 2% respectively after using the solution.
• Built calling strategy for the client using predictive modelling and k-means clustering
• This solution was built using Bayesian Augmented Network (TAN) algorithm but it was tested before on logistic regression, random forest, neural network- BPA and RBF.
• Social Media Network Risk Score for Consumer
• Developing a tool using social media and insurer’s data to build a counter fraud model for property & claims insurance vertical.
• This tool will comprise of social media network between customers, garage owners, medical professionals, and drivers, forecasting of provider claims, risk score modelling. It will also help investigation team in pre and post processing of claims
• Counter Fraud Model for Healthcare Insurance
• Developed a tool that will help client validate companies and hospitals by using their online information.
• This tool is built using data from geo-coding api for address validation, domain information, plagiarism on website content and images, social media information (twitter, LinkedIn and Facebook) about domain owner’s
• Text Mining
• Built a solution for open ended customer comments using LDA and Gibbs Algorithm, Hierarichal Clustering, Bayesian and text blob noun classification algorithm. The combination of algorithms is able to compute sentiments, classification, summarization and topic identification
• Fraud Risk Indicator
• Built an unsupervised learning algorithm by defining rules & guidelines related to claim amount, transactions related to withdrawal, name & address matching algorithm, and churning algorithm. This algorithm is able to identify suspicious agents and policies leading to fraud, link graph between fraudulent agent & policy holders.

Apr 2013 – Sep 2014 · 1 yr 5 mos · Gurgaon, India

• Medicare & Dual Growth Model
• Developed a business solution to predict and infer change in Medicare and dual market in post US healthcare reform scenario using different uptake assumptions based on wage and quartile index for a state.
• This solution was built using detailed business knowledge, demographics, and geographic factors involved in US at most granular level possible
• Market Share and Game Theory Simulators for leading US Healthcare Clients
• Developed market share prediction tool which incorporates demographic, socio-economic factors and product features in calculating market share for a particular company in various regions.
• This tool also computes the population that will be playing part in those regions. It also predicts the market share and their population segment wise
• It helps a company to understand how their product features designed is affecting their market share in various segments and by changing these features can increase their market share.
• Different statistical technologies and methodologies where used in building this project such as logistic, multi-nomial logistic, regression, poisson regression, mean absolute percent error, Price elasticity curve
• Time Zero LTV Models for leading US Healthcare client
• Developed time zero LTV/acquisition models based on axciom data to acquire new customers keeping new US healthcare reform guidelines using regression, generalized linear model, t-test
• Admission Readmission Model
• Predicted the percentage of in-hospital admissions for a company in order to reduce claims.
• This detailed study figured out that a small percent of hospital admissions can reduce major proportion of claims by assigning either appropriate product to customer or not retaining the customer
• Retention Models for leading US Healthcare Client
• Developed models in healthcare revenue solutions for product mapping and to increase annual margin of clients
• Different methodologies were used- Regression, Generalized Linear Model

Oct 2012 – Apr 2013 · 6 mos

• Development of campaigns tailored to the targeted customer based on input provided by clients in UNICA and SAS
• Analysis, programming, testing, Quality check, problem solving and documentation.
• Single point of contact for database and UNICA with respective administrators
• Development and automation of reports using SAS and tableau
• Plays a significant role in PMO team for automation of monthly metrics and campaigns.
• Training to new joiners and helping them understanding database and process quickly
• Counts Calculator
• The significance of this project is to bridge down the gap between technical and process management people. Development of code is based on different segment and demographics involved using SAS and minimal occupancy of database space
• Profiling Calculator
• Development of this project helps one to look at a particular segment and the percentage of customers available in that particular sub segment

Jun 2011 – Oct 2012 · 1 yr 4 mos

• Development of campaigns tailored to the targeted customer based on input provided by clients in UNICA and SAS
• Analysis, programming, testing, Quality check, problem solving and documentation.
• Single point of contact for database and UNICA with respective administrators
• Development and automation of reports using SAS and tableau
• Plays a significant role in PMO team for automation of monthly metrics and campaigns.
• Training to new joiners and helping them understanding database and process quickly
• Computing demographics based on online behavior of users
• Developed a statistical model which computes the number of users in a household and predict their demographics (age, gender and income) based on online behavior of single user in other household
• Different techniques were used during completion of this project such as canonical discriminant analysis, regression, principal component analysis in SAS and usage of R software to build regression tree.

May 2010 – Jul 2010 · 2 mos · Bangalore

• Designed and developed an innovative system which helps in alignment of the misaligned slices of the brain images. Implemented a novel technique of texture mapping to arrange these images into a 3Dimension stack which represents the structure of brain

Jun 2009 – Jul 2010 · 1 yr 1 mo · Gwalior Area, India

• Build car rental portal, different websites, research work and articles