Hey , I'm Ceren Kaya Akgün

Building AI automation workflows, LLM agents, and Retrieval-Augmented systems that connect real-world data sources with intelligent assistants. Discover my latest experiments in agentic orchestration, workflow automation, and applied generative AI.

About Me

With 7+ years across engineering, product, and analytics, I now build AI automation workflows that orchestrate LLM agents, vector pipelines, and operational integrations. That cross-functional experience helps me translate messy requirements into reliable AI systems.

I focus on Retrieval-Augmented Generation, autonomous agents, and event-driven automation using platforms like n8n, Supabase, and modern LLM stacks. My projects combine real-world data sources, safety guardrails, and monitoring so they can run in production—not just demos.

I document experiments on GitHub and write about applied generative AI, workflow design, and tool-building patterns on Medium.

If you'd like to collaborate on AI automation, agent design, or workflow reliability, feel free to reach out on LinkedIn.

Projects

AI & Machine Learning Applications

DocQuery RAG

Automated RAG pipeline using n8n: Connects Google Drive, Supabase vector DB, and OpenAI for intelligent document indexing and AI-powered chat with semantic search.

Key Features:

  • 🧩 Auto-ingestion pipeline from Google Drive
  • 🧠 RAG chat interface with OpenAI/OpenRouter
  • 🔎 Semantic search & reranking using Cohere
  • 🧰 Postgres-backed chat memory
  • 🔔 Slack alerts for monitoring

JobMatch AI

An intelligent agent-based tool powered by AI and NLP to find jobs that match your profile. Upload your CV (PDF/JSON) or enter manually, and get personalized job recommendations using semantic similarity. Built with LangChain agents, HuggingFace embeddings, and Adzuna API.

Key Features:

  • 🤖 Agent Architecture with modular tool-based system
  • 📄 Multi-format Input (PDF, JSON, or manual entry)
  • 🔍 Real-time Job Search powered by Adzuna API
  • 🎯 Smart Matching with embeddings and cosine similarity
  • 🌍 Multi-country Support (Germany, UK, Netherlands, and more)
  • 🔧 CLI Interface for easy command-line usage

Flavor Bot

An intelligent recipe recommendation chatbot that suggests recipes based on ingredients or dish preferences. Powered by Spoonacular API for rich recipe data, it features advanced semantic search and LLM-driven query understanding (Groq/Ollama) to handle complex dietary requirements and natural language requests.

Key Features:

  • 🔍 Semantic search using transformer models
  • 🤖 LLM-powered query understanding with Groq/Ollama
  • 🥗 Automatic dietary restriction & ingredient exclusion detection
  • 🌐 Dual Interface: Interactive Web UI & Command Line (CLI)
  • 🛡️ Enterprise-grade guardrails: Input validation & prompt injection protection
  • 🐳 Production-ready with Docker support & comprehensive logging

Quiz Generator

An AI-powered web application that automatically generates interactive quizzes from any web content using LlamaIndex and OpenAI. Features include URL-based quiz generation, multiple-choice questions, and real-time scoring.

Python Weather GUI

A weather application built with Python and Tkinter that provides real-time weather updates for any city. Features include temperature display, weather description, and a user-friendly interface. Built as a learning project to practice Python, GUI development, and API integration.

Final Assignment Agent (GAIA Benchmark)

This project is my submission for the Hugging Face Agents Course final assignment. I developed an AI agent that answers multi-step reasoning questions using the GAIA benchmark API. The agent retrieves questions, processes them, and submits answers for exact-match evaluation. The project demonstrates agentic reasoning, tool use, and API integration. You can review the project code and try the agent live.

Details: The assignment involves building an agent that interacts with the GAIA API to fetch questions, process them, and submit answers for scoring. The evaluation is based on exact match with ground truth answers. For more information about the assignment and the process, see the course details.

Data Science & Analysis

House Price Prediction

A comprehensive analysis of house prices using multiple machine learning models. Features include detailed feature engineering, model comparison, and performance analysis. Technologies used: Python, Scikit-learn, pandas, and various regression models.

Sleep Life Insights

A machine learning initiative that analyzes lifestyle factors and sleep patterns to forecast sleep quality. Features include prediction based on lifestyle factors, examination of sleep quality relationships, and interactive visualizations of sleep health metrics.

Customer Delivery Predictor

A machine learning model that predicts food delivery times based on various factors like weather conditions, traffic density, and location data. Features include real-time predictions, comprehensive feature analysis, and a RESTful API for easy integration.