Furkan Say
AI/ML & Backend Software Engineer · Adana, Türkiye
I am Furkan Say, a software engineer with 5+ years of backend experience focused on AI-augmented systems: Retrieval Augmented Generation (RAG), Large Language Model (LLM) agents, vector search with Qdrant, and polyglot microservices across C#, Python, Rust, Java, and TypeScript. I hold an M.Sc. in Computer Engineering with a thesis on deep learning and domain adaptation.
What I do
- AI / ML: RAG pipelines (hybrid dense + sparse retrieval, RRF, cross-encoder rerankers), LLM agents with tool use, PyTorch domain adaptation (DANN, BiLSTM-Attention, cWGAN-GP).
- Backend: .NET 8/10, ASP.NET Core, Clean Architecture, DDD, CQRS, MediatR, event-driven with RabbitMQ + Outbox, performance tuning (cut a critical API endpoint 85%, key SQL queries 60%).
- Polyglot: Spring Boot 3 (Java 21), FastAPI (Python), Axum (Rust), NestJS / Next.js 15 (TypeScript), gRPC, WebSocket/SSE streaming.
- Data: MSSQL, PostgreSQL (pgvector, row-level security), Redis, Qdrant, MinIO.
Selected projects
- Enterprise AI Assistant — multi-tenant RAG platform as a 5-language polyglot microservices monorepo (C#, Java, Rust, Python, TypeScript).
- Voice Notes TR — fully-local Turkish meeting-notes app (Tauri v2, faster-whisper, local Gemma 3 via Ollama).
- NexusPLC — real-time PLC monitoring (C#/WPF, Modbus, snap7, MQTT, Redis, RabbitMQ).
- IMU Behavior Recognition — M.Sc. thesis, PyTorch domain-adaptation pipeline (DANN, BiLSTM-Attention, cWGAN-GP).
Contact
This site is an interactive RAG-powered portfolio. Enable JavaScript to query a 3D embedding space of my profile and stream answers in real time.