I build practical AI systems around local LLMs, RAG pipelines, agent tooling, evaluation, and automation.
I am an AI engineer based in Türkiye, focused on turning LLM experiments into usable developer tools and product-like applications. My work sits at the intersection of retrieval systems, local-first AI agents, tool calling, evaluation, and MLOps.
|
Tool-using agents, local model workflows, reasoning traces, and agent diagnostics. |
Hybrid retrieval, reranking, document loaders, embeddings, and evaluation loops. |
Fast prototypes, clear documentation, tests, releases, and real user workflows. |
- Building explainable and self-healing AI agents
- Designing RAG libraries for JavaScript, React, and Python workflows
- Evaluating tool calling, retrieval quality, and local LLM behavior
- Working with Ollama, LM Studio, FastAPI, Flask, React, Docker, and vector stores
- Sharing open-source experiments, notebooks, and technical notes
|
Local-first framework for building explainable AI agents with self-healing, verbose traces, chaos testing, diagnostics, and multi-agent orchestration.
|
Production-oriented RAG toolkit for JavaScript, React, and Node.js with hybrid search, reranking, evaluation metrics, streaming, and local LLM support.
|
|
Memory-focused LLM experiments with quickstarts, graph memory demos, custom tools, streaming examples, security checks, and response metrics.
|
Flask-based assistant with session handling, SQLite persistence, document-aware flows, streaming chat, local embeddings, and UI improvements.
|
I like building projects that are easy to run locally, easy to inspect, and useful for other developers. My current open-source direction is centered on:
Local LLMs + Retrieval + Tools + Evaluation + Clean Developer Experience
- Local-first: Ollama / LM Studio compatible workflows
- Explainable: traces, diagnostics, scoring details, and failure analysis
- Practical: quickstarts, examples, tests, packaging, and release notes
- Composable: small libraries and templates that can be reused in real apps
|
|
“The best way to predict the future is to invent it.” – Alan Kay
Build · Test · Share