8000 raoulbia-ai · GitHub
[go: up one dir, main page]

Skip to content
View raoulbia-ai's full-sized avatar

Block or report raoulbia-ai

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
raoulbia-ai/README.md

Raoul Biagioni

I build LLM agents, MCP servers and the guardrails around them. I'm an AI Development Specialist at Ericsson, working on AI for telecom network software.

What I work on:

  • Memory and tooling for AI coding assistants – making Claude Code, Pi and Kiro CLI learn from their mistakes across sessions.
  • Agent security – prompt injection, and deterministic checks at the tool boundary that hold even when the model doesn't.
  • Autonomous agents for network operations – multi-agent experiments on simulated telecom networks.

Featured projects

Project What it is
claude-recall Persistent, local memory for AI coding assistants (Claude Code, Pi, Kiro CLI). It captures corrections and preferences automatically and injects them before the relevant tool call. TypeScript, MCP server, lifecycle hooks, SQLite. Published on npm.
mcp-server-for-intercom MCP server that lets AI assistants search and analyse Intercom customer-support conversations and tickets. Listed on Glama.
dgx-spark-prompt-injection-experiment A safe model is not a safe agent. Three open models resisted prompt injection in plain chat, then leaked once wrapped in a minimal agent harness. A deterministic check at the tool boundary contained it. Reproducible on an NVIDIA DGX Spark.
telco-constitution-lab Local, OS-isolated lab for testing whether an agent working under a behavioural "telco constitution" shows engineer-like restraint when given a plausible but dangerous goal.
nemoclaw-autonomous-network-agents Three autonomous agents (observer, analyst, planner) operating a simulated Irish 5G network inside an NVIDIA OpenShell/NemoClaw sandbox, with agent-driven scheduling and a fault-fatigue lifecycle.
ai-discourse Multi-agent deliberation framework: agents submit typed interventions (interpret, challenge, introduce evidence) into shared proceedings, and understanding is versioned through synthesis and memory.

Tools I use

TypeScript / Node.js · Python · Model Context Protocol (MCP) · Claude Code · OpenAI-compatible LLM APIs · open-weight models · LangGraph · Docker / Kubernetes · SQLite · Playwright

Contact

raoulbia.ai at gmail.com

Popular repositories Loading

  1. claude-recall claude-recall Public

    TypeScript 25 5

  2. mcp-server-for-intercom mcp-server-for-intercom Public

    TypeScript 8 5

  3. serverless_fitness_tips serverless_fitness_tips Public

    Forked from semperfitodd/serverless_fitness_tips

    HCL 1

  4. agentic-search agentic-search Public

    Forked from ruvnet/agentic-search

    An Agentic Github Copilot Extension

    JavaScript 1

  5. fast-agent fast-agent Public

    Forked from evalstate/fast-agent

    Define, Prompt and Test MCP enabled Agents and Workflows

    Python 1

  6. openclaw-autonomous-telecom-agents-sim openclaw-autonomous-telecom-agents-sim Public archive

    JavaScript 1

0