Backend engineer (Go, C# / .NET, TypeScript) who builds money-safe systems and obsesses over what breaks in production — idempotency, concurrency, exactly-once semantics, and failure recovery — all backed by real-Postgres tests, race-checked CI, and ~88% coverage. I've shipped a concurrency-safe double-entry ledger (gRPC, transactional outbox, no overdraft under load) and an ~8.6M-orders/s C++ matching engine. METU Computer Engineering with a national-level competitive-programming background. Looking for the correctness-critical backend work fintech teams hire for.
Guarantees exactly-once transfers and zero overdraft under concurrent load (row-level locking, sorted lock order) — proven by real-Postgres integration tests and a race-detector CI gate (~88% cov). gRPC surface, transactional outbox feeding an AML consumer, loan-origination FSM, and a dbt/DuckDB risk layer.
Header-only, zero-dependency engine with strict price-time priority (LIMIT/MARKET/IOC/FOK, O(1) cancel, L2 depth): ~8.6M orders/s single-threaded, hardened with property + fuzz tests (invariants + quantity conservation) and deterministic replay under ASan/UBSan CI.
ASP.NET Core gateway that stays up under pressure: rate limiting, request coalescing, a circuit breaker, and stale-cache fallback so a flaky upstream never cascades into an outage. Tests, Docker, and CI.
Idempotent Greenhouse→HRIS/Airtable provisioning so duplicate or out-of-order events never double-provision: HMAC webhooks, an ordered saga with retry/backoff + dead-lettering, drift reconciliation, and an audit trail. 22 tests, ~98% cov.
Tauri v2 desktop app that drives the Claude Code CLI headless and renders its agents, tool calls,
and code as live, navigable graphs — a Rust bridge streaming stream-json into a
vanilla-JS canvas reducer (pan/zoom board, lanes, replay).
Contributed to and maintained the PHP backend of ocenchik.ru — a long-running Russian valuation & appraisal (ocenka) portal with ~40K monthly visits and ~39K indexed pages — for 6 years: server-side logic, request handling, and a database-backed data layer, keeping a high-traffic production site reliable.
Built backend functionality and the structured datasets powering this English-learning platform (~36K monthly visits, ~9K indexed pages), including the data model behind its learning content.
Taught Python, Flask/FastAPI, C#, and Unity in individual and group classes; guided students through building real backend projects — fundamentals first.
For fun — an optimized mobile destruction-physics pipeline (Teardown-style voxel destruction) tuned hard for performance and memory, in a published title.