Ritvik Pandya is an engineering manager at JPMorgan Chase focused on merchant payments, distributed systems, and the reliability of financial infrastructure. His work combines learned execution graphs, deterministic routing, and event-driven merchant onboarding to reveal how transactions move through complex platforms and where failures emerge.
Earlier in his career, Pandya worked on enterprise applications and integrations across financial services and healthcare, including projects associated with Principal Financial Group, Discovery Health, and Quest Diagnostics. His professional portfolio also describes work associated with Apple and Google. At PayPal, he contributed to merchant onboarding and integrations involving Facebook Marketplace, Shopify, PrestaShop, Venmo Mastercard, and American Express Gateway. He subsequently moved into payments engineering leadership connected with WePay and JPMorgan Chase.
How he approaches payment infrastructure
- Execution graphs that pinpoint failures. Pandya represents each payment request as a short-lived directed acyclic graph spanning gateways, authentication, orchestration, downstream services, and notifications. This structure makes it possible to isolate a slow foreign-exchange service or distinguish a single failing component from system-wide degradation. His AI Engineer presentation describes combining OpenTelemetry, asynchronous Kafka processing, and tail-based sampling without slowing transaction processing.
- Baselines tailored to actual transactions. Domestic payments, cross-border transfers, individual clients, and different endpoints require different operational expectations. Pandya uses client-specific anomaly detection, Kullback–Leibler divergence, and exponential moving averages to reduce misleading alerts. He separates structural drift, rising traffic volume, and changing transaction composition because each demands a different remedy.
- Routing designed around data ownership. His work on deterministic routing and database locality examines how predictable request placement improves cache locality and limits coordination across application instances, while accounting for hot keys and uneven load.
- Onboarding as distributed coordination. His Kafka-based merchant-onboarding architecture separates identity verification, compliance, risk assessment, activation, and notifications into asynchronous stages. His work on self-healing ingress extends the same operational discipline to automated recovery: identify what changed, explain the proposed intervention, and control rollout or rollback before an incident spreads.
