I build production services for Simpo.ai across workflow orchestration, Kafka-backed microservices, CRM features, AI SEO tooling and real-time updates.
90% less manual pipeline work with metadata-driven Airflow DAG generation
80% lower Kafka consumer lag with per-tenant rate limiting
20+ shared tenants protected from high-throughput tenant spikes
Built a metadata-driven Airflow orchestration platform with Python, Kubernetes and custom operators, exposing workflow configuration APIs through Java, Spring Boot and microservices to dynamically generate cron-scheduled ETL DAGs and reduce manual pipeline development effort by 90%.
Designed per-tenant rate limiting and Kafka queue isolation across Java Spring Boot microservices, keeping high-volume jewelry retail tenants from impacting 20+ shared tenants and cutting Kafka consumer lag by 80% during peak traffic.
Integrated AI-based SEO keyword generation with OpenAI, producing industry-specific keyword suggestions grouped by low, medium and high competition.
Built a production CRM module for client records, meeting scheduling and real-time WebSocket notifications, with secure AWS S3 storage for documents and images.
Built real-time features with Kafka and WebSockets for instant UI updates and stream processing in the internal analytics pipeline.
Mentored and onboarded 3+ interns on Java, Spring Boot, microservices and development workflows until they contributed production-ready changes.
Aug 2024 to Oct 2024Proficonlabs Pvt Ltd. Hyderabad, Telangana
Software Intern
Cut the open bug backlog by 25% in 4 weeks across 6 Spring Boot microservices, fixing MongoDB query inefficiencies, Redis cache stampedes and REST contract mismatches.
Kept unit-test coverage above 80% with JUnit and Mockito, catching 3 regressions before they reached production.
JUnit, Mockito, MongoDB, Redis, REST APIs
How I build
Four habits behind the numbers above.
Isolate the noisy neighbour before it becomes an outage.
In multi-tenant systems, one customer's peak should never become everyone's incident. Limits, queues and budgets are set per tenant from day one.
Every action should be safe to run twice.
Idempotency keys, retries with backoff and dead-letter storage mean a failed payment or webhook is handled once and never lost.
Automate the repeated work, then measure the saved time.
Metadata-driven workflows, generated DAGs and observable services help teams move faster while still proving the platform is healthy.
Leave the codebase easier for the next person.
Test coverage above 80%, clear contracts between services, and real time spent helping new engineers ship on their own.
Toolkit
What I use day to day, grouped by where it sits in the stack.
Languages
JavaPythonJavaScriptTypeScriptSQL
Backend and frameworks
Spring BootMicroservicesREST APIsJPA / HibernategRPCSpring SecurityOAuth2JWT
Messaging and caching
Apache KafkaRedisWebSockets
Databases
PostgreSQLMySQLMongoDBCassandraChromaDB
Architecture patterns
Event sourcingIdempotencyCache-asideRate limitingPub-sub messagingREST API design
AI and LLM
LangChainRAGMCPVector databasesAzure OpenAITool callingAgentic AI