Emre.dev
AI Developer & Software Engineer

Cihat Emre Karataş

Local-first AI products, RAG systems, and intelligent web applications.

I turn LLM ideas into usable software, from memory-enabled assistants and internal chatbots to production-ready products built with Python, Django, React, and modern AI tooling.

View Case Studies Get in Touch

Selected work

Projects that show the direction

A smaller, stronger selection focused on local-first AI, internal assistants, RAG, and production-minded developer tooling.

View All Projects
Python · Local LLMs

Mem-LLM

A Python library for privacy-first, memory-enabled assistants that run on local LLMs with persistent conversation history and configurable knowledge bases.

  • • Persistent multi-user memory
  • • Pluggable storage and knowledge backends
  • • Designed for local Ollama workflows
GitHub
Internal Assistant

Kurum Asistan Chatbot

A local LLM-powered assistant for internal workflows, support requests, document uploads, and centralized organizational information.

  • • Internal knowledge access
  • • Multi-feature assistant workflow
  • • Local deployment focus
GitHub
JavaScript · RAG

Quick RAG

A production-oriented RAG toolkit for JavaScript and React apps with hybrid search, caching, conversation management, and evaluation.

  • • Frontend-friendly developer experience
  • • Hybrid retrieval and caching
  • • Evaluation-aware workflow
GitHub

Capabilities

What I Build

AI-focused software with enough product and engineering depth to move beyond the first demo.

AI Assistants & Agents

Tool-using assistants, local LLM workflows, memory systems, and automation around real tasks.

RAG & Knowledge Systems

Document understanding, retrieval pipelines, semantic search, and internal knowledge assistants.

Backend & APIs

Reliable APIs, authentication, databases, and scalable backend systems with Django and FastAPI.

Full-Stack Products

Responsive product interfaces and end-to-end web apps with React, Tailwind, and Python backends.

Writing

Notes from the build process

View Writing

AI Agents Need Control Systems, Not Just Better Models

The current AI conversation is shifting from model capability to operational control: agents need permissions, audit logs, kill switches, and domain-specific governance before they can safely act at scale.

Read note

OpenClaw and the Real Test for Personal AI Agents

OpenClaw is interesting because it exposes the real product problem behind personal AI: agents need memory, tools, permissions, and safety boundaries before they can become useful daily software.

Read note

Have an AI product idea worth building?

I’m open to roles, collaborations, and freelance projects around LLM applications, local-first AI, RAG systems, and intelligent web products.

Email Me Connect on LinkedIn