AI & RAG Researcher | LLM Fine-Tuning & Data Engineering | Open-Source Contributor
I am an AI Developer and Computer Science Researcher focused on Retrieval-Augmented Generation (RAG), Parameter-Efficient Fine-Tuning (PEFT/QLoRA), and Domain-Specific Dataset Curation. My work bridges the gap between raw institutional disclosures and context-grounded AI decision systems.
My primary research centers on solving numerical hallucinations, context adherence, and knowledge retrieval limitations in complex domains like higher education financial aid systems, NLP language learning tools, and web-scale information extraction.
My open-source models and benchmark datasets are hosted on Hugging Face:
| Asset Type | Repository & Link | Description |
|---|---|---|
| 🧠 Model | qwen2.5-scholarship-rag-adapter |
QLoRA fine-tuned adapter built on Qwen2.5-0.5B-Instruct optimized for higher education RAG pipelines. |
| 📊 Dataset | us-college-roi-scholarships |
Consolidated dataset mapping U.S. university financial aid, net price, and ROI outcomes. |
| 🧪 Dataset | learn-rag-eval-qa |
Synthetic Question-Answering dataset engineered for RAG evaluation (Faithfulness, Recall). |
learn-rag: Core RAG pipeline architecture featuring vector search, context ingestion, and fine-tuned LLM integration for domain query resolution.nanogpt-playground: Experiments with lightweight transformer architectures, causal language modeling, and tokenization mechanics.raft-lab: Research environment focused on Retrieval-Augmented Fine-Tuning (RAFT) techniques to improve model alignment with retrieved context.how-to-learn-ai: Curated roadmap and technical guides covering foundational AI, PyTorch workflows, and modern LLM engineering.
path-to-usa-scholarships: Institutional data collection, analytics, and ROI modeling for U.S. higher education financial disclosures.Climate-energy: Exploratory data analysis and predictive modeling centered on global climate metrics and energy transition trends.
ielts-grammar-studio: Interactive NLP tool for IELTS grammar assessment, error analysis, and essay scoring logic.SAT-Vocabulary/sat-vocab: Automated vocabulary acquisition systems tailored for standardized test preparation.PDF-quiz: Document processing system that extracts context from PDFs to generate dynamic evaluation quizzes.context-vocab-saver&Vocabulary-bank: Contextual vocabulary building utilities utilizing semantic tagging.Grammar-mastery: Rule-based and model-assisted grammar evaluation framework.
Virexa.Co&virexa-yade.premium-menu: Production-grade web architecture, UI/UX components, and dynamic menu engines.image-search-engine: Multi-modal or metadata-driven image indexing and retrieval tool.volunteer-hub: Community organization and volunteer management platform.denizli-guide&welcome-turkey: Regional information and localization applications.planlayici&focus-lab: Productivity and personal workflow management systems.
- AI/ML & LLMs: PyTorch, Hugging Face (Transformers, PEFT, Datasets), LangChain, QLoRA, Fine-Tuning
- Vector Search & IR: ChromaDB, FAISS, Semantic Search, Embedding Models
- Languages & Frameworks: Python, JavaScript, TypeScript, HTML/CSS
- Data & Tools: Pandas, NumPy, Git, Docker, Linux, Jupyter