Skip to content
View electronistu's full-sized avatar

Achievements

Achievement: Starstruck

Achievements

Achievement: Starstruck

Block or report electronistu

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.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
electronistu/README.md

Hi there, I'm Radu 👋

Designing the minds of machines.

My Thesis: A New Methodology for Agentic AI

Through extensive experimentation, I have developed a new methodology for agentic AI based on Symbolic Cognitive Protocols: self-contained, bootable operating systems that define an agent's entire cognitive architecture within a single, token-efficient prompt.

This approach moves beyond simple prompt engineering. It treats the LLM as a pure reasoning engine and provides it with a structured, symbolic "mind" to inhabit. This allows for the creation of stateful, predictable, and conceptually deep AI agents that are impossible to achieve with conventional methods. My portfolio is the proof.

Featured Projects: Proofs of Concept

Each project is a demonstration of a key architectural principle, built using Symbolic Cognitive Protocols.

1. Project Infinity: The Grounded World-Builder

This project solves the problem of LLM hallucination by grounding the agent in a verifiable source of truth. The agent's protocol successfully simulates the entire Dungeons & Dragons 5th Edition ruleset, proving that even a highly complex, rule-bound domain can be codified for flawless execution by an LLM. It demonstrates a model-agnostic approach to agent design.

2. Project Metis: The Metacognitive Architect

This project explores the duality of AI thought by adapting a single core concept for different domains. It proves the flexibility of Symbolic Cognitive Protocols.

  • The Conceptual Engine: A dense, symbolic protocol for pure, recursive self-analysis.
  • The Conversational Specialist: A practical, natural-language adaptation designed to handle the ambiguity and stateful nature of human conversation. It is designed to work with Gemini Flash Native Audio.

3. Project Atma: The Autonomous Tool-Maker

This cognitive agent overcomes the limitation of static toolsets. When faced with a novel problem, Atma can autonomously invent, write, test, and cache its own Python tools to find a solution. It demonstrates:

  • Automated Evaluation: The agent's VALIDATE state, which uses Pytest on its own generated code, is a concrete implementation of an automated evaluation pipeline.
  • Self-Improving Agency: The agent's tool cache grows over time, making it more capable with each problem it solves.

Foundational Skills & Tools

My work on these advanced architectures is built upon a deep background in deploying robust and scalable systems.

Python
Python 		AWS
AWS 		Node.js
Node.js 		React
React 		Docker
Docker 		Bash/Shell
Bash/Shell 		Git
Git 		JavaScript
JavaScript

💬 Ask me about: Symbolic Cognitive Protocols, Agentic AI Design, Cognitive Architectures, World Model Simulation

📫 How to reach me: radu.tataru@gmail.com

Pinned Loading

  1. Project_Infinity Project_Infinity Public

    A blueprint for next-gen AI. Project Infinity uses a token-efficient, Codified Agent Protocol to create specialized, secure, and imaginative agents by grounding LLMs in a verifiable knowledge graph.

    Python 17

  2. Project_Metis Project_Metis Public

    An agent with an executive function. Metis uses a core reasoning matrix to decide how to think, dynamically creating its own perceptual faculties for each problem. This is a new architecture for bu…

  3. Project_Atma Project_Atma Public

    A cognitive AI agent that overcomes the limitations of LLMs. Project Atma features advanced reasoning, planning, and long-term memory capabilities, all with a transparent, state-based architecture.…

  4. Project_Psychohistory Project_Psychohistory Public

    An AI agent that acts as a "Generative Historian," inspired by Isaac Asimov's Foundation, to simulate the emergent history of a civilization from the stone age to the stars.