Skip to content
View abutbul's full-sized avatar
50 followers · 106 following

Achievements

Achievement: StarstruckAchievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Achievements

Achievement: StarstruckAchievement: QuickdrawAchievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Highlights

  • Pro

Block or report abutbul

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
abutbul/README.md

Hi, I am David Abutbul — 4X5PI

AI/LLM Research · Open Source Advocate · Security-Minded Engineer · SecOps · Infrastructure · Automation · Radio Amateur 4X5PI

About

LLM security researcher and old-school BBS head. I build MCP servers, LLM agent tooling and retro systems. Running a Synchronet BBS for fun, keeping the 80s-90s scene alive. Give me a dial-up tone and a 16-color palette and I am home.

Experience

RoleCompanyPeriod
LLM Security ResearcherSentinelOneNov 2025 – Present
LLM Security ResearcherPrompt SecuritySep 2025 – Nov 2025
IT Specialist (Automation & Infrastructure)FinalOct 2022 – Aug 2024
DevOps EngineerDeep InstinctMar 2020 – Oct 2022
Solution Center ManagerDeep InstinctFeb 2018 – Mar 2020
SaaS Operations – Change Management Engineer / Application SupportRSA SecurityJul 2012 – Feb 2018
IT ManagerTradeFXL/Markets.comSep 2009 – May 2012
Sysadmin, HelpdeskBynet Data CommunicationsMay 2006 – Dec 2007
Role highlights

LLM Security Researcher · SentinelOne / Prompt Security
Sep 2025 – Present

  • Probe and stress-test large language models for vulnerabilities, exploits, and unintended behaviors.
  • Design red-team style attacks on LLM pipelines to uncover weak points before malicious actors do.

IT Specialist (Automation & Infrastructure) · Final
Oct 2022 – Aug 2024

  • Infrastructure Innovation: Designed IaC-driven systems (Ansible/Python) that cut deployment time and replaced legacy tools with open-source solutions.
  • Knowledge Sharing: Established a centralized documentation hub (Confluence/Wiki), slashing cross-team onboarding time.

DevOps Engineer · Deep Instinct
Mar 2020 – Oct 2022

  • CI Pipeline Optimization: Revamped Jenkins-based CI pipelines, reducing deployment cycles through automated testing and quality gates.
  • Devex and Dev Enablement: Shared Library management, Artifactory guidelines and infrastructure, Slack/GitHub integration.

Solution Center Manager · Deep Instinct
Feb 2018 – Mar 2020

  • Built and managed a team responsible for production and configuration management, including customer-specific setups (e.g., security thresholds, compliance rules) and maintenance of staging/production environments for AI-driven cybersecurity solutions.

SaaS Operations – Change Management Engineer / Application Support · RSA Security
Jul 2012 – Feb 2018

  • Accompanied the integration of new features from SRS to deployment, providing feedback to RND/QE teams to safeguard deployment in production and align with PCI requirements.
  • Developed scripts, guides and documentation for handling cryptographic keys.

Projects

ProjectWhat it doesStars
Prompt Security · Open-source work, contributed from my work account davida-ps.
clawsecSecurity skill suite for OpenClaw-family agents: drift detection, audits, integrity checks.
JavaScript Agents Security		1136
ps-fuzzMake your GenAI apps safe & secure: test & harden your system prompt.
Python LLM Fuzzing		720
RAG_Poisoning_POCStealthy prompt injection and poisoning in RAG systems via vector database embeddings.
Python RAG Security		37
rag-poisoning-workshopHands-on workshop: RAG poisoning / indirect prompt injection via vector DB embeddings.
JavaScript RAG Workshop		0
Personal
openapi-mcp-generatorTurn OpenAPI/Swagger specs into Docker-ready MCP servers (SSE/IO, auth).
Python MCP Docker		28
mod-game-state-apiAzerothCore game-state API.
C++ Game		6
NoDrumsKaraoke track maker using Spleeter + sox (vocal/drum separation).
Python Audio ML		5
docker-knapsack-llmComputer-use Docker player for LLM research.
Docker LLM Shell		4
mcp-hostAPI connecting an Ollama model to MCP servers for agents.
Python MCP Ollama		1
imgenGenerate consistent 4-directional sprite views.
Python Graphics		1
gizbarServerless, client-side expense tracker.
TypeScript		0
trbbsRule-based Telegram bot system.
Python Bots		0

Skills

LanguagesPython TypeScript JavaScript C++ Bash
DomainsLLM security & red teaming RAG poisoning & prompt injection MCP servers LLM agents BBS systems amateur radio game modding audio DSP
ToolsGit GitHub Actions Docker LangChain ChromaDB React FastAPI Ollama

Contact

GitHub abutbul · Web abutbul.github.io/4X5PI · LinkedIn in/abutbul

Banner drawn in 16-colour ANSI and generated from code — generator/generate.py · data/profile.json

Pinned Loading

  1. openapi-mcp-generator openapi-mcp-generator Public

    A Python tool that automatically converts OpenAPI(Swagger, ETAPI) compatible specifications into fully functional Model Context Protocol (MCP) servers. Generates Docker-ready implementations with s…

    Python 28 9

  2. docker-knapsack-llm docker-knapsack-llm Public

    Computer use Docker Player (LLM Research)

    Shell 4

  3. mcp_host_wrapper.py mcp_host_wrapper.py
    1 
    import traceback
    2 
    import logging
    3 
    import sys
    4 
    import os
    5 
    from langchain_mcp_adapters.client import MultiServerMCPClient
  4. mcp-host mcp-host Public

    A simple API that connects to an Ollama model and MCP server for agent interactions.

    Python 1

  5. mod-game-state-api mod-game-state-api Public

    AzerothCore game state api

    C++ 7 6

  6. prompt-security/clawsec prompt-security/clawsec Public

    A complete security skill suite for OpenClaw, Hermes, PicoClaw and NanoClaw agents (and variants). Protect your SOUL.md (etc') with drift detection, live security recommendations, automated audits,…

    JavaScript 1.1k 110