Backend Systems Engineer | Distributed Systems & Fintech
I design, build, and operate backend systems where data integrity, retries, and failure handling are non-negotiable. I specialize in building boring systems that don't break under load.
- Built and operated backend workflows supporting 100k+ paid subscribers in revenue-critical environments with strict consistency guarantees.
- Reduced infrastructure and API costs by $48k/year by migrating third-party workflows to in-house PostgreSQL systems over 6 months.
- Improved query performance by ~90% in production financial data pipeline serving 3M monthly users through schema redesign and index optimization.
- Reduced LLM token costs by ~30% using semantic caching and request-aware model routing.
FSM-based transactional workflow engine designed to prevent state corruption in payment and finance workflows.
- Enforces exactly-once state transitions using idempotency keys
- Uses the transactional outbox pattern for safe event publishing
- Built to survive retries, crashes, and duplicate events
Deterministic async ingestion pipeline for large-scale document processing.
- Idempotent vector indexing with failure-safe retries
- DLQ-backed recovery for rate limits and partial failures
- Designed for predictable cost and zero data loss
Core stack: Python, PostgreSQL, Kafka (production-validated in finance-critical systems)
System design: Event-driven architectures, idempotency patterns, failure recovery
Currently exploring: Go for high-throughput API services
LinkedIn: linkedin.com/in/winsongr
LeetCode: leetcode.com/u/winsongr