English | Türkçe
I build real-time systems for financial markets. For the last 2.5 years I've been developing InfoX, a mobile trading app for the BIST and US markets — owning features end to end, from architecture to production.
- Run 15+ .NET microservices serving live market data to thousands of concurrent users
- Built the real-time market data feed: WebSocket → RabbitMQ → SignalR, with automatic primary/fallback failover per trading session
- Handle market-open spikes of 100,000+ messages/second without loss (bounded queues, backpressure, .NET Channels)
- Built the full fundamental analysis module: Hangfire jobs pull data from KAP, TEFAS and the Central Bank; a Strategy-pattern parser turns bank, insurance and industrial statements into 1,000+ metrics
- Shipped an AI feature that translates US market news into Turkish (OpenAI), translated once and served from Redis to keep costs low
- Cut database load ~60% with Redis caching; Circuit Breaker + Retry so one failing call can't cascade
Languages & Frameworks
Messaging & Real-Time
Data & Caching
Cloud & DevOps
tickforge-go — Multi-Exchange L2 Order Book Engine
Builds a consistent Level-2 order book from live Binance and Coinbase feeds.
| Correctness | Snapshot bootstrap, gap detection, automatic resync |
| Concurrency | Race-free by design — one goroutine owns each book, readers see an immutable snapshot via an atomic pointer, verified with Go's race detector in CI |
| Performance | Hand-written allocation-free JSON scanner: 0 allocs/frame, ~8× faster than encoding/json |
| Measurement | Honest tick-to-process latency, p50/p99 via HdrHistogram, reproducible benchmarks |
| Delivery | Distroless Docker image, GitHub Actions CI |
The original .NET prototype lives in TickForge — the Go rewrite is where active development happens.
search-engine-service — Multi-Provider Search Engine (.NET 10)
A search service that ingests data from multiple content providers (JSON & XML), normalizes it, ranks results with a weighted scoring algorithm, and serves them over a REST API. Built as a company case study against mock provider APIs.
| Architecture | Clean Architecture (4 layers), CQRS with MediatR, JWT auth |
| Search | Elasticsearch with BM25 relevance, fuzzy matching, prefix & wildcard queries |
| Data | PostgreSQL (EF Core, code-first) + Redis cache via the Decorator pattern with key-prefix invalidation |
| Events | RabbitMQ + MassTransit — async cache invalidation and reindexing after sync |
| Quality | xUnit + Moq + FluentAssertions; one-command docker-compose bringing up all 5 services |
finance-uni-club — Startup & Finance Community Platform
Community platform I built and run end to end for a university startup & finance club — live in production at startupvefinanstoplulugu.com.
- Next.js 14 + TypeScript + Sanity CMS, built for SEO and performance
- Deployed on Vercel with CI/CD and automatic preview builds on every pull request
building: tickforge-go — pushing the hot path toward zero allocations
learning: market microstructure, exchange feed protocols (ITCH/SBE), Go internals
open_to: fintech · low-latency systems · backend engineering · remote"Willing to question defaults and rebuild from first principles."
