Murali Krishna Vaddadi

• Seeking assignments in Analytics to provide efficient and intelligent solutions that help businesses achieve desired results such as reduction of cost and increase in productivity.
• SUMMARY :A result-oriented Analytics Professional with 16+ years of IT experience along with 7+years into Analytics. Graduated from” Indian School of Business (ISB), Hyderabad” in Advance management programme in Business Analytics.
• Analytics professional having knowledge in domains of Automobile, Insurance, Banking, Telecommunications, Healthcare and Pensions. Proficient in statistical modelling in Forecasting/Predictive Analytics, NLP, LLMs, Auto-ML,Machine learning algorithms etc.

• Filed "3 Patents" for the work I did in the area of AI,NLP,Q&A mechanism for efficient user query management using open document search methodology. It is under process with Attorney for formal filing.
• Received an award by TG government as part of 28th Edition of HYSEA Innovation Summit Awards 2020, Jointly being organized by HYSEA, STPI and Govt of Telangana for the work I did on Covid19. Noted, Prominent personalities attended in the event includes, Sri K. T. Rama Rao, Hon'ble Minister for IT, Mr. C.P. Gurnani, CEO of Tech Mahindra, Sri Jayesh Ranjan, IAS, Principal Secretary, ITE&C Department, and Dr. Omkar Rai, DG, STPI were the guest of honors for the event

Link for the same:

https://www.linkedin.com/posts/murali-vaddadi-4145144b_28th-category-ai-activity-6730747774226329600-jtJX

• Filed 3 patents for the work I did on effective question answer mechanism using open document methodology in NLP area. Currently it is in progress with Patent Attorney.
• Built a Digital CEO chat bot which can answer any queries from the given set of documents using open document search methodology.
• Built an Auto ML system which can provide predictions and forecasting results to the customers based on provided data

NBA (Next Best Action) designer template and X-CAR (extended customer analytical records model) in customer decision hub (CDH):

• I have created Next best action designer template and xCAR model in customer decision hub predominantly for Insurance & Healthcare verticals for different releases of CDH, same is available in Market place. Also, created campaigns for NBA scenarios in sample applications across both verticals provide NBA offers for the identified customers.
• Prediabetes outbound campaign: I have created NBA offer to the Prediabetes patients with a certain qualifying criterion and segregated them into different risk categories and then provide offers to the identified customers.
• MTM (Medication therapy management) campaign offers to customers.
• Helped consultants’ team for building and extending xCAR component for HC customers.

Platform Bug prediction model to find out turnaround time prediction for product release version bugs: (Quality control team)

• Every product-based company whenever they want to rollout new features, they will be releasing a new version of the product. It is often observed that, due to some change in the architecture some bugs get created as part of release.
• Based on priority, severity few bugs generally will take longer time to provide a solution to the given problems. So, handling those bugs at critical release times will be a hectic and challenging task. Our bug prediction model will help the application teams, with the prediction of turnaround time values for each bug and gives an estimate of when that bug can be completed as per the past bugs history.
• This will also help product teams to manage their team members intelligently to handle multiple bugs in each time frame and can allocate more resources if required, based on the current predicted time completion values.

Identification of customers who can buy/subscribe Term-deposits in banking industry:

• Created a term deposit AI model with the help of H20 & Pega predictive studio, which will identify the customer who can buy/subscribe to term-deposit product. This model can be used by any banks, to identify potential customers who can buy term deposit product.

Patient readmission prediction model for Pega Healthcare (Pega Care Management):

• Patients with serious and chronic illnesses are treated in the hospital and then discharged. Unfortunately, according to multiple studies, up to 25% of these patients will be readmitted within 30 days to be treated again, often with less favorable outcomes.
• With a focus on value-based care, providers are trying to prevent unnecessary readmissions and improve patient care outcomes. Readmission can be significantly reduced by taking steps while the patient is still hospitalized, defining different actions during discharge, and taking steps post discharge to ensure compliance with home care regimens, readmissions risk prediction can require data about the specific patient’s recent care, their current condition, treatment, their home life and other risk factors from electronic medical records.
• AI models can use this information to provide a proactive assessment of their risk and notify clinicians while the patient is still hospitalized. AI can provide the reasons that will lead to readmission and provide recommendations for the types of treatments that are most likely to be successful given the patient’s history.

Multi label classification customer feedback model:

• We have built a multi label classification model which will take the customer feedback data as input and classify the data into different labels for example like complaint, comment, bug, meaningless, request, undetermined etc. This can be used in any service industry where customer feedback is taken in the form of text. This AI model can be used by any business which deals with customer feedback, to improve their service in respective products.
• I have also, built a word cloud which will help us to understand all the high-level issues (example: small room size, not hygiene, food was bad) in a single view so that business can take some actions on the high priority issues.

Entity extraction using NLP to find out risks scores of customers:

• One of the Pega FS application team wanted a solution, which can identify the people who are involved in antimoney laundering, fraud etc., with the help of Natural language processing while onboarding a customer into application.
• We have created an NLP model, which extracted the entities (Name, address, company registration number etc.) from different links/social media and able to match the extracted entities with a given person, provided a risk score based on given similarity. Also, we can validate that person/organization, in case if they were involved in any fraud/illegal activities.

Identity verification component for faster processing of KYC of customers. (Banking):

• Generally, most of the banks takes 2-3 days to complete the KYC of a customer post verification of all documents. It is time consuming and sometimes it might take longer time if the documents are not in proper shape.
• With the help of Identity verification component, we can compare a person detail and provide a result with in 5 minutes by comparing person image/identity document.
• Identity verification component takes two inputs, person image/selfie and person identity card like Aadhar, driving license, Passport and does comparison and provides a result in case if that person image matched with given identity proof.
• Tools: Python 3.6 and Django framework 2.2 for API creation, Open cv, Pega decisioning AI product.

Face recognition live using Webcam and IP camera:

• Earlier we have had multiple discussions on feasibility of implementing a face recognition system at the security doors in place of swipe in of the ID cards for all employees.
• With the help of image recognition AI models, we have devised a solution which will capture the person image using webcam/IP camera, then it will identify the person using LDAP authentication server will provide the employee details like Full name, Employee Id, Email id, Department etc. on the screen while entering the office.

Prediction model to find out the most probable customers that can buy fund products:

• Client: Store brand (Norway)
• Tools used: Azure ML, R, SQL server
• Built a predictive model which can identify the potential funds (free fund, Ask etc.) buying customers along with their probabilities (more likely to buy, likely, less likely) for fund products, which in turn improve the profits of the client by targeting the appropriate customers.

Key achievements:

• Improved the ROC value of the model from 83% to 89%.
• Improve the client profits by acquiring more fund customers.
• Helped the client to target the real buying customers instead of contacting all customers randomly.
• Proper utilization of the fund advisors time to talk to the relevant customers.

Prediction model to find out most probable customers for Insurance products (Health insurance, Child insurance):

• Created a stored procedure to get the consent counts from the customers to use their data for product recommendations/upsell/cross sell in view of GDPR implementation.
• Done a data quality analysis for the one of the third-party data sources and identified the logical inconsistencies in few scenarios and presented results to the client business team with findings and suggestions.

Prediction model for identification of fraudulent transactions in corporate transactions:

• Domain: Banking and Finance
• Client: Citi Bank
• Tools: R, Scala2.10, Hadoop, Spark1.6.2, Hive, Cloudera, Machine learning techniques

Key Achievements:

• Reduced false positive rate of fraudulent detection from 15% to 1% as per the client requirement.
• Improved accuracy of model performance to 96% by using ensemble techniques.
• This model helped the client in saving cost by detecting future frauds in the corporate transactions.

Defect prediction model for software release bugs:

• Domain: Manufacturing furniture products
• Client: DFS
• Tools: R, Hive DB

As part of this project, we have built a Defect prediction model which will forecast the no of defects for the next release.

Recommendations to the customer:

• This solution helped the client in saving the cost by planning/allocating the resources based on forecasted bugs for the next release.

Prediction model for finding outages in event data(TAPE):

• Domain: Healthcare
• Client: UHG
• Tools: R, SQL server

As part of the project, we built a classification model to classify the event as outage and non-outage. Also identified and listed out variables which may cause outages using Association rules.

Recommendations to the customer:

• Provided insight on highest priority tickets outage, application wise outage ticket details with an accuracy of 86%
• This solution helped the client with improved service delivery by reducing service disruptions.

Text analytics model to find themes in the mobile brands data:

• Domain: Telecommunications client
• Client: T Mobile
• Tools: R, Excel files

Developed a machine learning solution that will identify the different available brands in the given data and identifies different sentiments (Positive/Negative/Neutral) associates with those brands/services for one of the Top telecom clients.

Recommendations to the customer:

• Provided recommendation on what are the negatives about a particular brand so that it will be rectified by the client and improve their service quality.
• This solution helped the client in identifying the negative points and rectify them to provide the better service to the customers.

Prediction model to find out fraudulent transactions in the health care insurance claims:

• Client: Discover
• Tools used: R, Tableau

Recommendations to the customer:

• This solution helped client saving money on identifying and rejecting fraudulent insurance claims using advance AI/ML models.

Prediction model for loan repayment estimation of customers to predict good, bad, default customers:

• Tools used: SAS R miner, Tableau

Recommendations to the customer:

• As a solution insight to the client recommended different behavior statistics (low credit score, Usage or cards, high valued previous loans) of default customers as an insight.

 Pega certified Data scientist