Oğulcan AKCA

Multi-Agent system with Supervisor an Coordinator agent using LangChain, LangGraph, Gemini, ChromaDB, Streamlit.

A multimodal product recommendation system integrating LLM with RAG and GraphSAGE neural networks. Leverages semantic search, graph-based product relationships, and fashion domain expertise for intelligent recommendations. Built with PyTorch, ChromaDB, and Streamlit for interactive product discovery.

No description provided.

No description provided.

No description provided.

No description provided.

No description provided.

MCP server for AI agents to collect user input via dynamic web forms. Just provide field names, AI generates the UI. Supports SSE transport, Docker deployment, and OpenAI Agents SDK guardrails/tracing. Best for agentic workflows needing real user data.

A Personal Learning Assistant built with OpenAI Agents Python SDK. Features multi-agent architecture with handoffs, guardrails, context management, tracing, and SQLite sessions - following OpenAI's official SDK standards and best practices.

It aims to train and use a LoRA (Low-Rank Adaptation) model on Stable Diffusion v1.5 that can mimic the distinctive visual style of “Family Guy”.

No description provided.

This repository includes models that predict Amazon's stock prices using machine learning techniques and Docker integration with the REST API built with the best model.

İzmir toplu taşıma verilerine erişim sağlayan bir Model Bağlam Protokolü (MCP) sunucusu, AI asistanlarının şehir ulaşım verilerini ve analizlerini sorgulamasına olanak tanır.

Identify email addresses or domains names that belong to colleges or universities. Help automate the process of approving or rejecting academic discounts.

Pixel-level steel defect segmentation on the Severstal dataset using a U-Net model with an EfficientNetB4 backbone, implemented in TensorFlow/Keras.

An AI agent project built with FastAgent, leveraging Model Context Protocol (MCP) servers for seamless interaction with diverse services including Google Maps, Slack, Gmail, SQLite, screen capture, weather information, and web search. This project demonstrates a rich, interactive environment with comprehensive service connectivity for AI agents.

Agent-to-Agent (A2A) communication protocol for inter-agent coordination and a Model Context Protocol (MCP)-inspired architecture for interacting with external tool servers.

MIDI Music Generation Experiments with LSTM

Image Super Resolution with Deep Learning (Bicubic vs. SwinIR)

It is to automatically detect human faces in a given image and recognize their identities by comparing the detected faces with a predefined gallery of people. The system should also be able to label unrecognized faces as “Unknown”.

An Experiment to Improve Image Classification Performance Using Data Augmentation with GAN

It applies the Knowledge Distillation technique to transfer the knowledge of the BERT model (Teacher) to a significantly smaller Transformer model (Student). The aim is to reduce model size and potentially inference time, while maintaining high performance or improving on standard training.

A project where Text-to-Image, Image-to-Image, Inpainting, ControlNet, LoRA Loading features are implemented.

This repository includes models that predict Amazon's stock prices using machine learning techniques and Docker integration with the REST API built with the best model. In addition this features, AWS EC2 server was integrated.

ChatGPT 4o model was used to develop question-answering systems using RAG technique with the generated website log dataset.

This is the final course in the IBM Data Science Professional Certificate as well as the Applied Data Science. This capstone project course has gave me the chance to practice the work that data scientists do in real life when working with datasets. In this course I've assume the role of a Data Scientist.

A study was conducted to develop different recommender systems on a three-year customer-product relationship data. SimpleSVD, NormalSVD, MidUserSVD, MidItemSVD, MidTotalSVD, KNN, KMeans, RFM analysis, MBA and Apriori techniques were used.

It is an application to generate similarity marking data that can be given as input to a machine learning model in Python using the Flask library.

This is a Python script that creates a dashboard using the Dash and Plotly libraries. The dashboard displays performance data for domestic flights in the United States. I developed this project using skills I learned from the IBM Data Science Certificate program.

I created it using multiple datasets and the skills I learned from the IBM Data Science Certificate program. The project involves scraping data from various websites. With this tool, users can extract information from the web quickly and easily. The code utilizes various libraries such as BeautifulSoup and Requests.