At John Wiley & Sons, failed and missed eCommerce orders once took approximately three months to detect. Venkata Ramana Saddi applied machine learning to transaction reconciliation, helping reduce that interval from 3 months to a few hours. Business stakeholders confirmed the improvement, which enabled teams to identify affected orders and prioritize immediate reprocessing, particularly for high-value transactions.

The result reflects an area of Saddi's expertise that connects machine learning with enterprise integration: finding discrepancies across business processes and making those findings useful to the people responsible for resolving them. His subsequent work at Chubb brought transaction pattern matching into billing reconciliation, alongside underwriting modernization and notes summarizer delivered.

Through 2024, his professional experience spanned nearly two decades across publishing, banking and insurance. As PRS Technology Leader at Chubb and previously Director of Software Engineering at Wiley, he combined hands-on architecture and implementation with responsibility for engineering delivery.

How Transaction Pattern Matching Improved Wiley's eCommerce Operations
Saddi applied a transaction pattern matching algorithm to reconciliation, connecting the matching approach with the movement of information across Wiley's enterprise applications. Digital order reconciliation went live in late 2018, during his consulting engagement supporting Wiley.

Salesforce eCommerce order reconciliation followed in early 2020, and reconciliation for product and pricing records was introduced in late 2021.

For Wiley's operations team, earlier detection meant they could identify failed or missed orders sooner and decide which needed immediate attention. High-value orders could then be prioritized for reprocessing, giving the business a practical way to act on the issues Saddi's reconciliation work uncovered.

His contribution depended on understanding the business workflow as well as the integration architecture. The successive releases extended reconciliation from order transactions to product and pricing records, applying transaction matching to distinct parts of Wiley's commerce environment.

Bringing Transaction Matching into Chubb Billing
In 2023, Saddi designed an event-driven billing reconciliation framework at Chubb and implemented a transaction pattern matching algorithm using his machine learning expertise. The framework connected billing and related application flows to identify missing transactions and support investigation. Detection time fell from days to minutes.

The project placed machine learning within a production business workflow. Saddi combined the transaction-matching approach with integration design so that discrepancies could be surfaced for operational review. The work demonstrates the application of his reconciliation expertise in both publishing commerce and insurance, with each implementation addressing its own business context.

Improving the Underwriting Platform
Saddi led architecture and delivery for underwriting capabilities covering policy inquiry, policy notes, referral modernization and client profiles. This work improved access to information across cloud and mainframe systems while preserving connections to applications still supporting business operations.

For the policy-notes migration, he designed the Azure SQL data model and migration utility for 28.2 million notes and 350 million lines of text from a legacy IMS database. The redesigned approach combined historical and incremental data loads, reducing the migration window from approximately 55 days to 2.5 days. He also led the first Personal Risk Services disaster recovery test on an Azure SQL database.

Saddi also led performance optimization for the cloud-mainframe integrated policy inquiry application. Confirmed 2025 results included performance gains of 5x to 27x across measured workloads, response-time reductions of 85 to 96 percent and 99 percent fewer failures. A written assessment by Chubb's PRS chief information officer credits his redesign with these improvements and describes the change from three months of daily outages to six months without repeat disruptions.

Referral modernization introduced reusable APIs and a cloud-hosted rule engine. Rule-change cycles fell from more than six months to one day. Phase 1 was simplified from 145 rules to 55 while retaining functionality, making an established business capability easier to maintain and change.

Applying Generative AI to underwriting Notes
In 2024, Saddi implemented a notes summarizer using generative AI. The capability summarized claim information and answered users' questions about the notes, helping underwriting users understand the material and locate relevant details during review. His approach emphasizes human review and accountability, with summaries and answers supporting the people evaluating the underlying claim information.

Research in Fraud Detection and Machine Learning
Saddi's published research adds a scholarly dimension to his applied work. His 2024 articles include "Real-Time Insurance Fraud Detection Using Reinforcement Learning" and "Leveraging Big Data and AI for Predictive Analysis in Insurance Fraud Detection." His 2023 publications include "Role of Natural Language Processing in Detecting Insurance Fraud" and "Insurance Fraud Prediction: An Analysis Using Supervised and Unsupervised Machine Learning."

His wider research covers hybrid AI/ML fraud detection, generative AI for cybersecurity and deep learning in medical imaging. His conference experience includes the 2023 IEEE Technology & Engineering Management Conference and other international technical conferences in 2023 and 2024. In 2012, he presented a code-analysis framework for integration tools at a TIBCO technology conference.

Developing Shared Engineering Capabilities
Saddi also translated project experience into shared engineering practices. He authored more than 25 shared API specifications and helped establish coding, delivery-pipeline and disaster recovery standards adopted across PRS teams. He led design forums and coached engineers in profiling and root-cause analysis, extending the work beyond individual application releases.