Core Banking Modernization for Meridian Capital Bank

The Problem

Meridian Capital Bank, a regional bank with roughly $4.2 billion in assets, was running its core ledger on a COBOL mainframe system first deployed in the early 1990s. Every account balance update, interest accrual, and regulatory report depended on an overnight batch job that took nearly six hours to run, meaning transactions posted the following business day rather than in real time. The bank's few remaining engineers fluent in the mainframe codebase were nearing retirement, and every new product — from a promotional savings rate to a new overdraft rule — required months of change requests to a vendor with no meaningful competitive alternative.

Customer expectations had moved on. Meridian's mobile app showed stale balances, real-time fraud alerts were impossible on a next-day processing cycle, and integrating with modern account-aggregation services like Plaid required brittle, custom workarounds. Leadership had shelved two prior modernization attempts after vendors proposed multi-year, high-risk "big bang" cutover plans that the board considered too risky for a system holding customer deposits.

The Solution

TechNova proposed a phased strangler-fig migration instead of a big-bang rewrite: a new event-driven ledger, built on Kafka and Spring Boot microservices, ran in parallel with the mainframe, first mirroring transactions for validation before gradually taking over posting responsibility account-segment by account-segment. This let Meridian de-risk the migration in production with real transaction volume rather than staging-environment assumptions, and gave the risk and compliance teams a continuous audit trail comparing old and new system output during the transition.

Each account migration batch ran through automated reconciliation tooling we built specifically for this project, comparing balances, transaction history, and interest calculations line-by-line between systems before customers were cut over — any discrepancy above a defined threshold halted the batch automatically rather than risking a customer-visible error. The new platform was deployed on Kubernetes on AWS, provisioned entirely through Terraform, with a blue/green deployment pipeline that let Meridian ship changes weekly instead of the quarterly release windows the mainframe vendor required.

On top of the new ledger, we built a digital banking API layer used by Meridian's mobile and web teams, real-time fraud-signal streaming replacing the old next-day batch fraud report, and Plaid-based account aggregation so customers could finally see external accounts alongside their Meridian balances.

Results

The migration ran for fourteen months with zero unplanned customer-facing downtime, moving all 340,000 customer accounts off the mainframe in twelve incremental batches. Nightly batch processing — previously a six-hour window that constrained everything from same-day transfers to customer support response times — dropped to eleven minutes, and transaction posting moved from next-day to real time. Fraud detection, which had relied on the prior day's batch output, now flags suspicious activity within seconds.

With the mainframe fully decommissioned, Meridian eliminated $3.6 million in annual licensing and support costs and, more strategically, cut the time to launch a new deposit product from months to weeks — turning what had been a multi-year technical liability into a platform the bank's product team now builds on with confidence.