Backend Engineer · Java · Distributed Systems · Event-Driven Architecture
I build backend systems where correctness under failure matters: message ordering, retries, idempotency, recovery, concurrency, and transactional consistency. Currently a Software Development Engineer at Kaleris, working on N4, a terminal operating system used by major container ports.
- Cut replica rebuild time from ~28 to ~11 minutes for ~800K entities by parallelizing and batching state hydration and restart recovery.
- Hardened message ordering across Kafka and ActiveMQ at ~1.5M messages/day, reducing ordering-related data defects from ~7 to ~2 per quarter without serializing processing.
- Reduced redundant downstream updates by ~25% with per-transaction coalescing and post-commit event publishing.
- Cut manual data-repair tickets from ~11 to ~3 per month using optimistic versioning, DLQ/redelivery, and compensating reconciliation.
- Built a minimum-commitment billing workflow across terminal operations and billing for ~150 vessel visits/month, from design review through UAT.
| Project | Pull Request | Contribution | Status |
|---|---|---|---|
| Spring Tools | #2055 | Built type-based @ModelAttribute validation with configurable severity, JDT quick fixes, and VS Code support. 1,100+ lines, 86 tests. 5.5.0 milestone. |
Merged |
| Apache HertzBeat | #4458 | Fixed bulk pause failing past Tomcat's request-size limit by moving IDs to a JSON body across Angular and backend, with backward compatibility and 15 E2E tests. | Merged |
| Keycloak | #53839 | Fixed SAML core null handling in XMLTimeUtil, where an error was created but never thrown. Regression tests added; 129 SAML Core tests passing. |
Merged |
| Spring Tools | #2049 | Fixed false-positive @ModelAttribute diagnostics that could silently change Spring MVC binding behavior. 5.5.0 milestone. |
Merged |
| Apache Fineract | #189 | Return HTTP 400 instead of 500 for invalid request payloads, with field-level validation errors and regression tests. | In review |
FairDrop — High-concurrency fair allocation platform
Java 21 / Spring Boot flash-sale allocation service with FIFO fairness and zero overselling via PostgreSQL row locking. Sustains ~800 requests/s at p99 ~150 ms across 5,000 concurrent buyers. Transactional outbox to Kafka, idempotent consumers, DLQ, Redis rate limiting, 57 concurrency/integration tests, and Micrometer/Prometheus observability.
| Platform | Achievement |
|---|---|
| LeetCode | Guardian |
| Codeforces | Candidate Master |
| AtCoder | Blue-rated |
| Meta Hacker Cup | Round 2 qualifier (2025) |
- Languages: Java, SQL, JavaScript, TypeScript, C++, Python
- Backend: Spring Boot, Spring Integration, Spring MVC, Hibernate/JPA, REST APIs
- Messaging: Apache Kafka, ActiveMQ/JMS, Transactional Outbox, Idempotent Consumers
- Data: PostgreSQL, SQL Server, Redis, Flyway
- Infrastructure: Docker, Kubernetes, Google Cloud (Professional Cloud Architect), Prometheus, Grafana