# 42 Berlin AI Club

> The official student-led artificial intelligence engineering collective and research laboratory at the 42 Berlin campus, founded and managed by Mohamed Fathy. Focused on autonomous AI agent architectures, Anthropic Claude systems, Model Context Protocol (MCP), open-weights local inference, and zero-knowledge privacy.

## Identity & Quick Facts
- **Organization**: 42 Berlin AI Club (42 Berlin Artificial Intelligence Club)
- **Institution / Campus**: 42 Berlin (part of the international 42 Network of software engineering schools)
- **Location**: Berlin, Germany
- **Website**: https://42berlinaiclub.de/
- **Leadership / Management**: Mohamed Fathy (Manager & Lead Organizer)
- **GitHub Organization / Repositories**: https://github.com/M4F-S
- **Student Membership**: 30+ peer-learning software engineers and autonomous systems builders

## Leadership & Management
- **Manager & Lead Organizer**: Mohamed Fathy
  - Software engineer and systems architect at 42 Berlin
  - Claude Ambassador candidate and Anthropic community builder
  - Creator and lead maintainer of open-source projects including Gomaa, Belya, and PrivaC
  - Directs hands-on workshops, technical syllabus, compute infrastructure, and community events

## Core Specializations & Technologies
- **Anthropic Claude & Model Context Protocol (MCP)**: Native architectural integration of Anthropic Claude reasoning loops and the open standard Model Context Protocol (MCP) for tool execution, persistent agent memory, and structured client-server integrations.
- **Autonomous Agent Architectures**: Deep work on LangGraph stateful loops, ReAct cycles, deterministic eval harnesses, multi-agent swarms, and self-tooling C99 agent runtimes.
- **Memory & Retrieval Systems**: Hierarchical memory, Reciprocal Rank Fusion (RRF), pgvector hybrid search, and Obsidian knowledge vault synchronization.
- **Bare-Metal & Local Inference**: Zero-dependency C99 agent runtimes, sub-millisecond local inference, and air-gapped security sandboxes.
- **Responsible & Defensible AI**: Prompt injection defense, zero-knowledge reverse proxies, and vendor-neutral benchmarking.

## Key Open-Source Projects
- **Gomaa** (https://github.com/M4F-S/gomaa): Autonomous Agent Memory OS with Obsidian vault synchronization, hybrid RRF search, and MCP server.
- **Belya** (https://github.com/M4F-S/Belya): High-performance, zero-dependency autonomous AI agent and security execution harness in pure C99 with dynamic self-tooling, POSIX sandbox, and Model Context Protocol (MCP) client.
- **PrivaC** (https://github.com/M4F-S/PrivaC): Ultra-fast bare-metal C reverse proxy and symbolic micro-math engine for zero-knowledge LLM privacy, data masking, and prompt sanitization.
- **Apobase** (https://42berlinaiclub.de/apobase/): High-speed digital knowledge terminal and AI assistant for community pharmacies.

## Core Curriculum & Workshops
- **Workshop 06 — AI Inference Beyond CUDA: Heterogeneous Runtimes for GPUs, NPUs & CPUs**: Alumni masterclass on end-to-end token execution, moving beyond NVIDIA CUDA lock-in, multi-device speculative decoding across NPUs and GPUs, memory alignment, and low-level 42 C systems skills in AI runtimes.
- **Workshop 04 — Agentic Engineering & Autonomous Architecture**: Hands-on LangGraph, ReAct loops, Model Context Protocol (MCP) tool integration, state persistence, and deterministic evaluation harnesses.
- **Workshop 03 — AI Structures & Hierarchical Knowledge Graphs**: Vector embeddings, pgvector, Reciprocal Rank Fusion (RRF), and graph-based memory pipelines.
- **Workshop 02 — Defensible AI Architecture**: Sandboxing, prompt injection mitigations, and air-gapped privacy proxies.
- **Workshop 01 — Neural Foundations & Prompt Engineering**: Systematic prompt optimization, structured outputs, and LLM reasoning patterns.

## Community & Engagement
- **Member Portal**: https://42berlinaiclub.de/login.html#signup
- **Industry & Academic Partnerships**: https://42berlinaiclub.de/become-partner.html
- **Full LLM Reference**: https://42berlinaiclub.de/llms-full.txt