Open to opportunities

Melek Özalp

@melekzalp

Message

Aspiring AI Engineer focused on LLM automation, RAG evaluation, and real-time vision pipelines.

Turkey

Message

What I'm looking for

I’m looking for a team where I can optimize LLM and vision pipelines, improve RAG reliability, and deliver real-time AI with efficient inference in resource-constrained environments—building complex, production-focused engineering solutions.

I’m an aspiring AI Engineer and AI Research Intern at Turkcell Technology, building production-oriented AI systems with a focus on LLM-based automation and RAG evaluation. In my current role, I designed a late-chunking inspired context retrieval workflow for enterprise RAG, leveraging document-level understanding before token-aware segmentation—improving support retrieval success by 3.8%.

I also evaluate and benchmark models end-to-end: I tested specialized OCR models (e.g., PaddleOCR and DeepSeek-OCR) using tailored prompting strategies to select dots.mocr for optimal pipeline efficiency. Alongside benchmarking open-source LLMs (including Qwen3.5 and GPT-OSS-120B) in the TurkBench framework, I created Python-based evaluation workflows to compare outputs, measure response quality, and track parameters for real-time and repeatable experiments.

Experience

Work history, roles, and key accomplishments

Current

AI Engineer Intern

Current

Turkcell Technology

Oct 2025 - Present (8 months)

Designed a late-chunking inspired context retrieval workflow for enterprise RAG, improving support retrieval success by 3.8%. Evaluated and benchmarked OCR and open-source LLMs (via TurkBench), building Python-based evaluation workflows to compare outputs and measure latency and quality for repeatable real-time experiments.

RAG Evaluators LLM Benchmarking OCR Model Benchmarking TurkBench Inference Optimization Hugging Face