Backend Systems Engineer | Fintech | Scalable Architectures

Engineering resilient core systems for modern fintech.

Building high-performance, distributed systems that handle millions of requests at OneCard. Specialized in low-latency infrastructure and financial ledger integrity. Based in Pune, India.

Latency Optimization
45%

Reduced P99 latency across core payment gateways through JVM tuning and caching strategies.

Daily Volume
10M+

Managing transactions with strict ACID compliance and zero-loss reconciliation pipelines.

Throughput
50k+

System designed to scale up to 50k+ QPS during peak holiday traffic cycles.

Professional
Execution

Selected Works (2021-Present)

OneCard (Fintech)

SDE 2 | PUNE
  • terminal

    Led migration to Go-based microservices, achieving a 3x improvement in throughput and 60% reduction in memory footprint compared to legacy systems.

  • database

    Architected a real-time fraud detection bridge processing 5k+ events per second using Kafka and Flink with sub-50ms processing time.

  • security

    Redesigned the RBAC system to handle granular permissions for 100+ internal services, ensuring zero downtime during the migration phase.

Engineering Artifacts

hub
Go Redis

Distributed Rate Limiter

A sliding-window rate limiter designed for high-availability clusters. Solves the 'thundering herd' problem using Redis Lua scripts for atomic operations.

// Trade-off: Precision vs Latency v1.2.0

Used Token Bucket for smooth traffic shaping over Leaky Bucket to avoid request bursts during high-load intervals.

account_balance
Java Kafka

Payment Reconciliation Engine

Automated the matching of 1M+ bank settlement records daily. Handles multi-currency scenarios and partial payment exceptions with a 99.9% match rate.

// Architecture Snippet PROD_ENV

Idempotent consumers ensure no double-processing of settlement files. Implemented using Saga pattern for distributed consistency.

Case Study: 001

Designing a Low-Latency Ledger Service

The Challenge

How to maintain double-entry bookkeeping consistency for 50k transactions per second while maintaining a sub-20ms response time for balance queries.

The Architecture
ARCH_DIAGRAM_PLACEHOLDER // CQRS + Event Sourcing

Separated Read and Write paths. Used PostgreSQL for the source of truth (Append-only Ledger) and Redis for materialized balance views.

Performance Win

By moving from direct relational lookups to pre-computed materialized views in Redis, we achieved a 95% reduction in query latency during peak periods.

Languages

Java (Spring Boot) Go (Golang) Python

Infrastructure

AWS (EKS, RDS) Kubernetes Kafka Terraform

Databases

PostgreSQL Redis Cassandra
Technical abstract hardware view

Engineering Mindset

I believe in the First Principles approach to problem-solving. My work isn't just about shipping code; it's about evaluating architectural trade-offs, ensuring future scalability, and taking complete ownership of the system's lifecycle.

verified
Clean Code Advocate

Focusing on maintainability and readability as much as performance.

architecture
Trade-off Analysis

Understanding when to choose consistency over availability (CAP theorem).