Sep 2021 – Feb 2025 · 3 yrs 5 mos · Bangalore Urban, Karnataka, India · Remote

• Chargebee is a global leader in subscription billing and revenue management. I oversee the Data Platform, ML Platform, and our analytics solution, RevenueStory. We’re in the process of developing an Enterprise Data Platform designed to collect, organize, and process data securely, enabling both internal teams and our merchants to gain insights and make data-driven decisions. Additionally, we are creating machine learning models for smart dunning, customer churn prediction, revenue forecasting.

May 2019 – Sep 2021 · 2 yrs 4 mos · Bengaluru, Karnataka, India

• I served as the Head of Data Engineering and Data Science, where I built a comprehensive enterprise data platform capable of capturing and analyzing nearly a billion events and entities in near-real time from multiple data sources each day. At this stage, the compressed size of our data lake approached several petabytes. I also deployed various machine learning services and models, including Smart Routing to identify the optimal payment path (gateway, terminal) for maximizing transaction success, and Anomaly Detection to recognize unusual behavior in KPI metrics across various entity-dimension combinations, generating real-time actionable alerts.

Jul 2017 – Sep 2017 · 2 mos

• Customer Behavior Analysis and prediction:
• We will uses cluster analysis, decision tree algorithms to sub-divide each lifecycle stage into what we call segmentation layers. These layers group together the customers within each lifecycle stage who are similar to one another in terms of behavioral parameters, including spend level, activity frequency and lifetime value. Identifying homogeneous groups of customers based on their behavior is a powerful way to personalize customer communications for greatest impact.
• Prediction for best time, best channel for best ROI for customer engagement:
• Estimating best time to contact a customer in best channel for best ROI, includes estimating a statistical model which computes a score for determining a successful contact with the customer for the time period based on a first set of historical customer contact data among all connected channels like Email, SMS, POS, Social Media etc.. A second set of historical customer contact data associated with at least one customer may be received and the score of a successful contact may be provided for the customer based on the second set of historical data and the estimated statistical model.

May 2017 – May 2019 · 2 yrs

• FirstHive is a Customer Data Platform utilizing patent-pending technology to create unique customer identities by aggregating data from various customer interaction and transaction sources. It seamlessly integrates with all brand touchpoints, consolidating data into a single interface to develop rich customer profiles and intelligent cohorts. Its advanced customer journey orchestration capabilities allow marketers to leverage behavioral triggers from one channel to initiate actions in another.
• Role & Responsibilities:
• Architecture Responsibilities:
• Designed the overall architecture for the FirstHive product.
• Developed customer-centric models.
• Implemented Big Data architecture on Azure, utilizing HDInsight, Machine Learning, StreamSets, and Event Hubs.
• Established messaging architecture using ActiveMQ.
• Conducted social media analytics.
• Analyzed multi-channel marketing effectiveness.
• Data Science Responsibilities:
• Developed algorithms for unification and classification of customer data.
• Created narrowcasting algorithms.
• Conducted clickstream analysis.
• Analyzed multi-channel marketing responses.
• Performed sentiment analysis.
• Built a unified customer profile.
• Implemented collaborative filtering algorithms.
• Applied machine learning techniques for predictive analysis.
• Management Responsibilities:
• Collaborated with business stakeholders to create product vision and drive strategic direction.
• Ensured adherence to software engineering lifecycle best practices.
• Identified architectural weaknesses and recommended solutions in partnership with team managers.
• Actively participated in hiring and retaining top talent.
• Estimated efforts, planned and executed Scrum processes, identified roadblocks, and ensured timely delivery.
• Provided technical guidance, career development, mentoring, and constructive feedback to team members, including managers.

Apr 2017 – Sep 2017 · 5 mos

• WorldSwipe (www.WorldSwipe.com) allows shoppers to add your approved Loyalty wallets to Worldswipe, Manage all your points and spend from a single interface and Explore a host of rewards, privileges, and targetted offers, specially trailor made for you through Mobile App on website.
• WorldSwipe is an independent Value Exchange platform that can be plugged into a program created or run on FirstHive
• WorldSwipe will have access to data from multiple Accounts/Programs, who have enabled WorldSwipe catalogs in their respective Programs / Accounts
• No personally identifiable information will be accessible with any role on WorldSwipe (only data segmentation counts)
• The WorldSwipe ecosystem will have multiple roles which will have access to send campaigns to opt-in programs (e.g. category owner, campaign manager, etc.)
• WorldSwipe will also offer Customer to add self managed catalogs (as part of WorldSwipe Marketplace offers, available to only that specific program/account as a redemption option)

Apr 2016 – Apr 2017 · 1 yr

• Uniqification - An Machine Learning algorithm
• Our algorithms uniquified the interaction data collection from varies sources using statistical and unquification characteristics of a large social/data science dataset to better determined same users arrived in system from different sources but all user/social data collected may not be same.
• The Uniqification system is based on different Machine Learning and data science techniques.
• The system is using following algorithms to Uniqified the large social/data science datasets.
• 1.    Classification algorithm based on Machine Learning – The classification parameters are based Supervised Learning and divided to High, Medium and low confidence. The system is identify parameters based on Supervised Learning in training data. Many other classification parameters can be used based on Unsupervised Learning on user provided data.
• 2.    Customized Clustering algorithm – The Customized Clustering is used to  partitions a data set  into  homogeneous  groups based on  classification parameters such that similar  data sets  are  kept  in  a   group  whereas  dissimilar  data sets  are  in  different  groups.
• 3.    Unquification algorithms based on Positive Probability distribution – The data sets are grouped and classified as highly unique if Probability is high, classified as related data sets if Probability is Medium.
• 4.    Negative Probability impact on Unquification process - The probability of the outcome of an experiment is never negative, but quasiperiodic distributions can be defined that allow a negative probability for some unlikely events. These distributions has apply to unobservable events or conditional probabilities

Jan 2016 – Mar 2016 · 2 mos

• FirstHive, a SaaS based technology products platform, works with data and brings substantial impact in the key area of integrated marketing management strategy. FirstHive is using high invention to provide analytics for its customers.
• Architecture and design work:
• Integrated the historical transactional data and online ad campaign data
• Scored channels based on a performance index determined by new acquisitions, revenue and number of transactions
• Search engine optimization efforts by identifying essential location and behaviour based keywords
• Unearthed myths about customer channel preferences and helped formulate a customized marketing strategy around touch-points
• Benchmarked customers based on key marketing metrics
• Sentiment analysis - offers powerful business intelligence to enhance the customer experience, revitalize a brand, and gain competitive advantage.
• A 360-degree customer view offers a deeper understanding of customer behaviour and motivations.
• Ad-hoc analysis only looks at the data requested or needed, providing another layer of analysis for data sets that are becoming larger and more varied.
• Real-time analytics for quickly decipher and analyse data sets, providing results even as data is being generated and collected.
• Multi-channel marketing creates a seamless experiences across different types of media like websites, social media, and physical stores.
• Customer micro-segmentation provides more tailored and targeted messaging for smaller groups.
• Ad fraud detection requires data analysis of current fraud strategies by recognizing patterns and behaviours.
• Clickstream analysis helps to improve the user experience by analysing customer behaviour, optimizing websites, and offering better insight into customer segments.

Jan 2014 – Jun 2015 · 1 yr 5 mos

• Worked on Advance Topics of Big data Analytics and Virtualization Solutions as listed below:
• 1. Big Data Integrations
• 2. Metadata generation on unstructured big Data
• 3. PIG, Hadoop-R Analytics
• 4. Prediction Analysis, Forecasting Analysis for Business & System
• 5. Memory Based Change detection algorithms like Memory Based Graph Theoretic (MB-GT) & Memory Based Cumulative SUM (MB-CUSUM)
• 6. Designed Virtual Box Teradata virtual machine clustering for Linux server
• 7. Bonita Open Source research for Data Process Management of UDA eco-system
• 8. Architecting the Mongodb clustering virtualization solution for Teradata DBS Query Grid team & Teradata EVM Automation
• 9. Automated the cluster and configure the Mongodb VMs through puppet scripting
• 10. Research on Comparison matrix between Chef, Puppet, Docker and Vagrant
• 11. Provided Solution/POC for Over Provisioning issue of Cloud using Data & Decision Analytics. 12. POC to convert cluster with a 12:1 over provisioning to 4:1 over provisioning with good applications performance
• 13. Good In hand experiences on Hadoop Eco-system components like PIG, Sqoop, Hive, HBase, Oozie, Flume, Loom, Nutch and R

Jan 2013 – Jan 2014 · 1 yr

• Project #1: Proposing the use of sentiment analysis techniques on users’ personal text archives to aid in the task of personal reflection and analysis. This proposed system that processes an email archive, and slices it across different sentiment facets, such as those expressing various emotions, congratulatory messages, and messages related to family matters, religion, health, positive and negative sentiment of document/mail based on score.
• Project #2: Providing the REST architecture for NoSQL interface to database and enables application developers to write applications using a popular JSON-oriented query language created by MongoDB to interact with data stored of Teradata database through HTTP REST API calls. This driver based solution embraces the flexibility of the JSON data representation within the context of a RDBMS with well-known enterprise features and quality of service. This makes NoSQL solution to RestFull.
• Project #3: Generated models to for Throughput prediction in Teradata. Below Models predict the overall throughput of the system and identify the bottleneck resource.
• Each model produces an estimate of resource utilization at a given TPS rate. To determine maximum system throughput, identifed the TPS at which each model predicts the resource to be saturated For disk and CPU, we determined per-resource saturation points using the methodology for measuring the maximum resource capacity of a machine.

Aug 2012 – Jan 2013 · 5 mos

• Worked with leading e-commerce Company to create the POC on Big data analytic:
• The goal is to understand the e-commerce business from his websites and translate this information into actionable insights along with business performance and challenges.
• Comparison access multiple e-commerce sites and provide the missing products using web crawling.
• Worked on e-Commerce personalized offers platform. Enables merchant to track user’s behaviour and connect the dots to determine the most effective ways to convert one-time customers into repeat buyers.
• Worked on Price optimization using data science.
• Click path analysis on e-commerce site
• Market Basket Analysis - Created basket-level dynamics to make better decisions related to base and promotional pricing including
• Improve cross-selling opportunities across categories
• Up-sell to better or more profitable brands within purchased categories
• Build the holistic impact of promotions and price changes
• Improved performance of multiple-purchase offers

Feb 2012 – Jun 2015 · 3 yrs 4 mos

• Advanced Development Concept at Teradata R&D focuses on researching new ideas and areas, evolving them into development projects as needed. The insights gained from this research are then applied to Teradata Software Engineering, potentially impacting Enterprise Applications and Unified Data Architecture solutions.
• Key Projects in Research and Architecture:
• Teradata Virtual Edition/Machine: Visualization of the Teradata Database.
• Big Data Process Management: Streamlining processes for large-scale data handling.
• Excel Plug-in for Teradata: Facilitating data export and import.
• Teradata Compression Wizard: Enhancing data storage efficiency.
• Teradata NoSQL RESTful Driver: Enabling RESTful access to NoSQL data.
• Teradata-MongoDB REST Interface: Providing NoSQL support through a listener.
• Quantum TDCH: A unified TDCH Java API for Hadoop export/import.
• Map-Reduce Algorithms: Analyzing big data directly from Hadoop DFS.
• TPG BIRT Reporting: Reporting capabilities based on Eclipse.
• Teradata Capacity Planning and Prediction Analysis: Optimizing resource allocation.
• TVME Production Database Configuration: Designing for maximum database capacity.
• Integrated Solution for Teradata Index Tool: Enhancing indexing capabilities.
• Workload Balancing Techniques: Managing skew in Hadoop DFS for big data analytics.
• I was also focused on researching and architecting a data platform for various data servers (Hadoop, Oracle, DB2, Greenplum, MongoDB, Teradata, etc.). This architecture improves the common data server interface through an enabling layer of technologies (REST, OData) that supports SQL/file access, data loading and movement, RESTful and listener interfaces, and data and process management, emphasizing REST technology with standard data access interfaces like OData.

Mar 2006 – Aug 2012 · 6 yrs 5 mos

• I have built client development/sustaining team for Database Management, Active Support Management, Teradata Manager, Analyst Pack applications very quickly and absorbed all offshore activities at Teradata, Hyderabad India. My contribution was very high on completing complex RFCs and customer efixes in Active Support Management client applications (TDWM, TWA), Database Management and Database Query Analysis tools. I have demonstrated my skills on fixing DRs with less turnaround time. I have contributed heavily on four TTU main releases (TD 12.0, TD 13.0, TD 13.10 and TD 14.0).
• I have also contributed in many high priorities customers DRs in TASM area and fixed very quickly, make the team productive by leaning by doing methodology. I and my team have adopted Agile Project Management methodology first time in origination.
• I have worked part of TIES quality Warriors and provided many ideas which improved TIES DBS/Client development and testing capabilities, process and knowledge sharing.
• I have handled and developed following Teradata applications:
• Teradata Active System Management (TWA, TDWM) is a Goal-oriented, Automatic Management and Advisement technology in support of performance tuning, workload management, capacity planning, and configuration and system health management. Teradata ASM is capable of supporting complex workloads by adjusting via rules to various types of workloads.
• Teradata Manager (Dashboard and Trend Analysis): Teradata Manager Dashboard and Trend Analysis features proactively monitor the system through a workload-centric dashboard. It also provides various ad-hoc and standard reporting on workload behavior and trends.
• Teradata Analyst provides Teradata customers, both external and internal, tools for automating and easing the task of query performance analysis. Additionally, these tools help automate the difficult task of tuning the Active Data Warehouse to achieve better workload performance.

Mar 2004 – May 2006 · 2 yrs 2 mos

• Worked on all phases of data warehouse development life cycle for banking domain, from gathering requirements to testing, implementation, and support.
• Data modeling, responsible for gathering and translating business requirements into detailed, production-level technical specifications, creating robust data models, data analysis features and enhancements.
• Exceptional background in analysis, design, development, customization, and implementation and testing of software applications and products.
• Demonstrated expertise utilizing ETL tools, Data Transformation Services , and DataStage in Teradata.
• Strong leader with experience training developers and advising technical groups on ETL best practices.
• Excellent technical and analytical skills with clear understanding of design goals of ER modeling for OLTP and dimension modeling for OLAP.

Jul 1998 – Jan 2004 · 5 yrs 6 mos · Hyderabad Area, India

• Responsible for Database Query Analysis Tools design and development for Teradata Client Tools and Utilities organization of NCR using VC++/C++/Perl.
• Workload Profiler:
• The overall goal of Teradata Workload Management is to assist the DBA in configuring Teradata to provide the desired resource allocation for each workload. The assumption is that the workload requirements match or exceed the capacity of the system and the available resources need to be allocated based on the priority of the workload. Workload Profiler uses the collected information in the Teradata system and analyzes it to produce recommendations for workload definitions and classifications.
• Teradata Index Wizard (TIWIZ):
• Teradata Index Wizard is performance-tuning application that examines workloads, conducts analysis of the Indexes and suggests Indexes on these objects with the objective of improving the response time of the SQLs, thereby increasing the performance of the SQLs.
• Visual Explain and Compare Tool (VEComp):
• The purpose of VEComp tool is to visually depict the query execution plan generated by Teradata RDBMS Optimizer on the GUI and to resolve Optimizer-related discrepancies by visually comparing two executions plans.
• Teradata System Emulation Tool (TSET):
• The Teradata System Emulation Tool (TSET) provides an easy way to capture the source environment of a Teradata production system and emulate it on a Teradata test system of any size.
• TSET ensures that all elements of the production environment (for example, configurations, databases, data models and dependencies) are kept in sync, and all the required data is captured and successfully loaded onto the test system in the required format.
• SQL Sentence Generator (SQL-Gen):
• Sentence Generator Tool generates a variety of SQL statements by exercising Teradata SQL Grammar productions in a controlled fashion.