Skip to content
View mertdurukan's full-sized avatar

Block or report mertdurukan

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mertdurukan/README.md

Mert Durukan

Lead AI Engineer — LLM platforms · agentic systems · production MLOps
Ankara, Türkiye · remote-friendly

LinkedIn Email Medium


Six years taking enterprise AI from prototype to monitored production — LLM platforms, agentic systems, and applied ML. I ship the systems and lead the strategy, delivery, and teams around them.

Recent work

Stack

LLM & Agentic: LangGraph, LangChain, RAG (hybrid + rerank), RAGAS, LLM-as-judge, MCP, Langfuse, pgvector, Qdrant

ML / DL: Python, PyTorch, TensorFlow, scikit-learn, Vision Transformers, CNNs, YOLO

Platform & MLOps: AWS (EC2, S3, Lambda, RDS), Docker, Kubernetes, FastAPI, CI/CD, on-prem deployment


Reach out via LinkedIn or mert@durukan.dev

Pinned Loading

  1. agentic-rag-assistant agentic-rag-assistant Public

    Production-grade RAG over arXiv AI/ML papers — LangGraph, hybrid retrieval + cross-encoder re-rank, RAGAS eval, Langfuse observability.

    Python

  2. agent-eval-mcp-server agent-eval-mcp-server Public

    Production-grade MCP server that exposes AI agent quality evaluation (RAGAS faithfulness, precision, recall, response relevancy + custom JSON trace eval).

    Python

  3. imbalance-calibration imbalance-calibration Public

    Pre-registered study: class-imbalance corrections don't improve discrimination in tabular ML, wreck calibration, and the one apparent gain is an early-stopping artifact.

    Python

  4. mlx-quantization-calibration mlx-quantization-calibration Public

    Pre-registered study of how MLX quantization (bf16 → 2-bit, mxfp4, mixed recipes) breaks LLM confidence calibration — not just accuracy.

    Python

  5. tooth-resorption-detection tooth-resorption-detection Public

    Attention-augmented Vision Transformers for external root resorption detection on panoramic dental radiographs — improving macro-F1 from 0.20 to 0.90 over a CNN baseline. MSc Thesis, Mersin Univers…

    Python