Event-Driven and API-First: How Modern Architecture Unlocked $1.2 Billion in Student Loan Originations

Shalini Mani, a Senior Director at Discover, led the transformation of Discover Student Loans' digital platform. This article details how her API-first, event-driven architecture replaced legacy monoliths, resulting in a remarkable $100 million annual conversion uplift and setting a new benchmark for fintech modernization in the lending industry. Discover her strategic approach to innovation.

Total student loan debt in the United States now exceeds $1.84 trillion, spread across more than 42.8 million borrowers. Private student lending alone accounts for over $167 billion of that balance, and borrowers and cosigners collectively took on an estimated $102.6 billion in new loans during the 2024-25 academic year. Behind every one of those origins sits a technology stack responsible for verifying identity, pulling credit, validating income, and disbursing funds, often within a regulatory window that leaves little room for error. The platforms that power these workflows were, in many cases, built a decade or more ago, long before real-time processing and API-driven design became standard expectations. For lenders still running on legacy monoliths, the cost of that technical debt is measured not just in maintenance budgets but in lost applications, slower decisions, and borrowers who abandon the process before they ever receive funding.

Shalini Mani, a Senior Director at Discover with over two decades of experience spanning product management, financial services integration, and digital transformation, has spent much of her career confronting exactly this kind of complexity. A published author on HackerNoon, where her article "Designing for Regulation: A Fintech PM's Perspective" examines how regulated financial products can be built for speed without sacrificing compliance, Mani led the end-to-end modernization of Discover Student Loans' digital application platform in 2022. The project decomposed a legacy monolith into an API-first, event-driven architecture and delivered a $100 million annual conversion uplift, supporting more than $1.2 billion in yearly originations. Her approach treated the effort not as a technology migration but as a full product and business transformation, anchoring every architectural decision to a single principle: reduce friction for the borrower while preserving credit quality for the lender.

The Weight of Monolithic Systems in Financial Services

Financial institutions today spend nearly 40% of their IT budgets maintaining legacy platforms, and the average institution loses an estimated $93.6 million per year to inefficiencies rooted in outdated infrastructure. By 2028, banks and lenders that fail to modernize are projected to forfeit more than $57 billion collectively, with missed revenue in payments alone reaching 42%. The problem is structural. Monolithic architectures bundle every function, from credit checks to document generation to disbursement workflows, into a single, tightly coupled codebase. Changing one component means testing and redeploying the entire system, a process that can take months in a regulated environment where every release requires compliance review.

At Discover Student Loans, Mani encountered this constraint firsthand. The existing platform processed applications through a series of sequential, batch-oriented steps that could not scale independently. A spike in applications during peak enrollment season slowed the entire pipeline, and any modification to a single workflow required coordinating across the full stack. Mani's first move was to map every application drop-off and system bottleneck, identifying where borrowers abandoned the process and where internal latency created unnecessary delays. The analysis revealed that the architecture itself, not any individual feature, was the primary barrier to growth.

"When I looked at the data, the biggest drop-offs weren't happening because of poor UI or confusing forms," Mani explains. "They were happening because the system couldn't respond fast enough. A borrower would submit their information, and instead of getting a decision in seconds, they'd be waiting, sometimes for days. In a world where people expect instant answers, that delay was costing us millions in lost originations."

Decomposing the Monolith Into Independently Deployable Services

Financial institutions that have adopted microservices architectures report deployment frequency improvements of up to 61% and time-to-market reductions averaging 53% compared to monolithic approaches. A recent Gartner study found that nearly three-quarters of organizations are now using microservices in some capacity, and those in financial services have documented a 42% improvement in their ability to respond to shifting market conditions and customer demands. The shift from monolith to microservices, however, is not simply a matter of splitting code into smaller pieces. In regulated lending, each service must maintain its own compliance posture, and the communication between services must preserve data integrity across every transaction.

Mani's team decomposed the Discover Student Loans platform into independently deployable services organized around core business capabilities. Identity verification, credit decisioning, data validation, and loan disbursement each became their own service, communicating through well-defined APIs and an event-driven messaging layer. This meant that a change to the identity verification workflow could be developed, tested, and deployed without touching the credit engine or the disbursement pipeline. Each service could also scale independently, so peak-season traffic on the application intake would no longer bottleneck downstream processes like school certification or income verification.

"The API-first approach forced us to think in contracts," Mani notes. "Every service had to clearly define what it expected and what it returned. That discipline eliminated a lot of the ambiguity that causes integration failures in complex systems. And because the services were event-driven, we could process steps asynchronously. A borrower didn't have to wait for one check to finish before the next one began."

Preserving Credit Quality Through Phased Migration

The global digital lending market reached approximately $566 billion in 2026, with AI-driven underwriting processes now controlling over 43% of that market and boosting approval rates by an estimated 25% without increasing risk. For lenders operating in regulated categories like student loans, speed of origination matters, but so does accuracy. A faster platform that approves the wrong borrowers or misclassifies risk is worse than a slow one that gets the numbers right. The challenge for any modernization effort in this space is maintaining the integrity of credit models and compliance frameworks while fundamentally restructuring the technology underneath them.

Mani addressed this through a migration strategy built on three engineering practices: feature flags, parallel runs, and A/B testing. Rather than replacing the legacy system wholesale, her team rolled out new microservices incrementally, running them alongside the existing platform and comparing outputs in real time. If the new credit decisioning service and the legacy system returned different results for the same application, the discrepancy was flagged for review before the new service went live. Feature flags allowed the team to enable or disable specific capabilities at the user level, controlling exposure and minimizing risk. This disciplined approach earned Mani Discover's President's Award, a recognition given to the top 1% of employees, which she has received twice during her tenure for delivering high-impact results in technically demanding programs.

"We never turned off the old system until we were confident the new one matched or exceeded its accuracy," Mani reflects. "The parallel runs were non-negotiable. In student lending, a wrong credit decision doesn't just affect the company. It affects a student's ability to pay for their education. We ran thousands of side-by-side comparisons before we migrated a single real borrower."

What Event-Driven Architecture Means for the Next Generation of Lending

The digital lending platform market grew from $10.6 billion in 2024 to $12.1 billion in 2025 and is projected to reach $24.1 billion by 2030, driven by the convergence of API-driven architectures, cloud infrastructure, and AI-powered risk modeling. Embedded finance, where lending services integrate directly into non-financial platforms and consumer applications, is expected to reach $251.5 billion by 2029, signaling a future where loan origination happens not just on a lender's website but inside the apps and workflows borrowers already use. For institutions that have already modernized their core platforms, this shift creates an opportunity. For those still running monoliths, it creates an increasingly urgent problem.

Mani's work at Discover sits at the intersection of these trends. The API-first, event-driven platform she built is, by design, composable. Its services can be called by internal channels, partner integrations, or future embedded-lending interfaces without requiring a new build for each use case. Having since moved into a role directing merger integration efforts for the combined Capital One and Discover entity, where she helps orchestrate the unification of over $250 billion in credit card and lending portfolios, Mani continues to apply the same architectural philosophy: modular design, phased execution, and relentless measurement.

"The best architecture is the one you can extend without rewriting," Mani observes. "What we built for student loans wasn't just a platform. It was a pattern. Decompose the monolith, define clean contracts between services, test relentlessly, and let the data tell you when you're ready to go live. That discipline applies whether you're processing loan applications or integrating two of the largest financial companies in the country."

From Conversion Uplift to Measurable Business Scale

European and UK banks report spending 4.7 times more on regulatory compliance when running legacy systems compared to modern alternatives, and roughly 80% of financial institutions worldwide still operate core banking applications built on COBOL, a programming language that dates back to the 1950s. Cloud spending in financial services has grown at an annual rate of 16.2% in recent years, reaching $77.9 billion, as institutions recognize that modernization is not optional but an operational imperative. For Discover Student Loans, the business case was validated by the results. The modernized platform delivered a $100 million annual conversion uplift, directly supporting more than $1.2 billion in yearly originations.

The improvement came from multiple vectors. Faster processing meant fewer borrowers abandoned their applications mid-stream. Real-time validation caught errors and missing information immediately, rather than surfacing them days later through a batch process. Event-driven notifications kept borrowers and cosigners informed at every stage, reducing anxiety and the support calls that come with it. Mani's work also caught the attention of the broader information systems community. She was recently selected as a peer reviewer for the Americas Conference on Information Systems (AMCIS) 2026, evaluating research submissions at one of the most recognized academic conferences in the IS discipline, a role that reflects the rigor she brings to both industry practice and scholarly inquiry.

"The $100 million number gets attention, and it should, because it represents real borrowers who completed the process and got funded," Mani says. "But the metric I'm most proud of is the reduction in time-to-decision. We took a process that used to stretch over days and compressed it to seconds for the initial credit check. That's the kind of change that earns trust."