ÖMER FARUK KOÇ

I work in artificial intelligence, machine learning, and deep learning, focusing on modeling, data preprocessing, and system development. My experience includes local model training, Retrieval-Augmented Generation (RAG), GPU optimization, and MLOps, with an emphasis on building practical and scalable AI solutions.

Legal AI Assistant: Processes thousands of legal documents with OCR + semantic chunking and integrates them into a RAG pipeline. Provides instant answers to user queries with relevant laws and case precedents; pipeline auto-updates when new documents are added.

LLM-Powered Dashboard: This LLM-powered dashboard analyzes uploaded Excel/CSV datasets, generating meaningful summaries and visualizations. Users can interact with the data through natural language queries, receiving instant insights and actionable reports. Designed to make data-driven decision-making accessible to non-technical users.

Brain MRI Tumor Detection with Transfer Learning: Built a high-accuracy brain tumor classification system using EfficientNetB3 with Mixup augmentation, label smoothing, and a 2-seed ensemble. Achieved ~99.5% test accuracy with explainable results.

CLTV Modeling with BG-NBD & Gamma-Gamma: Developed a Customer Lifetime Value (CLTV) model using BG-NBD and Gamma-Gamma frameworks. Modeled purchase frequency and monetary value to generate actionable insights for marketing strategies.

CIFAR-10 CNN with Mixup & Label Smoothing: Implemented a CNN-based image classifier on CIFAR-10 with Mixup augmentation and Label Smoothing to reduce overfitting. Achieved robust generalization across classes with improved accuracy.

Amazon Reviews Sentiment-Based Ranking: Designed a sentiment-driven ranking system for Amazon reviews. Applied time-weighted average scoring and Wilson Lower Bound to surface the most reliable reviews, improving recommendation quality.

Machine Learning