Flynn Cruse | Open Source Interests

(919) 592-2092

flynn@netfree.net

UK/US Dual Citizen

https://www.linkedin.com/in/flynncruse/

Open Source Project Interests

A collection of open source projects that I've set up, analyzed, and studied extensively. These cutting-edge AI research frameworks have significantly influenced my work and thinking in AI, agent systems, and technical development.

The implementation insights gained from working with these projects have directly informed architectural decisions in our education platform development and other professional initiatives.

Disclaimer: These are projects I have explored, studied, or otherwise found valuable in my professional development.

Featured Open Source Projects

Below are the projects I've highlighted on my resume, along with dozens more that have been influential in my professional development. The first three projects represent cutting-edge AI research frameworks I've particularly focused on.

SakanaAI / AI-Scientist

Description: The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery (Lu, Chris and Lu, Cong and Lange, Robert Tjarko and Foerster, Jakob and Clune, Jeff and Ha, David).

A groundbreaking system for fully autonomous scientific discovery using frontier LLMs—by running end-to-end experiments across generative AI, diffusion models, and grokking. My own thinking about novel agent design and multi-modal data generation was refined through exploring this project.

Citation: arXiv:2408.06292

ShengranHu/ADAS

Description: Automated Design of Agentic Systems (Hu, Shengran and Lu, Cong and Clune, Jeff).

An eye-opening project that explores recursive AI design through a "meta-agent" capable of inventing and coding novel agents—pushing the frontier of emergent agency, and LLM-based invention. Working with ADAS gave me firsthand insight into the pace and potential of AI-driven agentic system design, and deepened my understanding of emergent agency, self-improvement, and the broader implications for AI development. It sharpened my perspective on both the possibilities and the responsibilities that come with increasingly autonomous systems.

Citation: arXiv:2408.08435

Genesis-Embodied-AI/Genesis

Description: A Universal and Generative Physics Engine for Robotics and Beyond.

A unified physics and generative agent framework for automating multimodal data generation—spanning physical video realism, camera and character motion, robotic control, interactive 3D scenes, articulated object generation, and expressive speech animation.

Repository: GitHub Repository

Agno

Description: A lightweight library for building multimodal AI agents that can generate text, image, audio, and video. It provides a unified API for LLMs with enhanced capabilities like memory, knowledge integration, tools, and reasoning.

Key Features:

  • Lightning-fast agent creation (10,000x faster than LangGraph according to their claims)
  • Model-agnostic design supporting any provider without lock-in
  • Native multimodal support for text, image, audio, and video
  • Multi-agent capabilities for specialized team collaboration
  • Memory management for persistent agent sessions

Technology Stack: Python, Various LLM providers, Vector databases

Fastbook

Description: Educational materials for deep learning based on the book "Deep Learning for Coders with Fastai and PyTorch" by Jeremy Howard and Sylvain Gugger. Provides comprehensive Jupyter notebooks covering deep learning concepts and implementation.

Key Components:

  • Complete curriculum covering deep learning from basics to advanced topics
  • Implementation examples using fastai library and PyTorch
  • Chapters on computer vision, NLP, tabular data, collaborative filtering
  • Architecture deep-dives into CNNs, ResNet, and optimization techniques

Technology Stack: Python, fastai, PyTorch, Jupyter Notebooks

GitDiagram

Description: A tool that transforms GitHub repositories into interactive system design/architecture diagrams for visualization. It allows users to quickly understand repository structure and navigate to relevant files.

Key Features:

  • Instant visualization of any GitHub repository structure
  • Interactive components with direct links to source files
  • Fast generation using AI (Originally Claude 3.5 Sonnet, now OpenAI o3-mini)
  • Customizable diagrams with regeneration options

Technology Stack: Frontend: Next.js, TypeScript, Tailwind CSS, ShadCN; Backend: FastAPI, Python, Server Actions; Database: PostgreSQL with Drizzle ORM; Deployment: Vercel (Frontend), EC2 (Backend)

ACE Framework (Autonomous Cognitive Entity)

Description: A comprehensive framework for building autonomous AI agents with their own cognitive architecture, memory systems, and ability to pursue objectives independently.

Key Applications:

  • Personal Assistant/Companion: Self-contained AI interacting with one user
  • Game World NPCs: Autonomous characters with personalities, motivations, and memories
  • Autonomous Employee: AI designed to perform productive work within corporations
  • Embodied Robot: Framework for self-contained, autonomous machines

Technology Stack: Python with various LLM integrations

Self-Operating Computer Framework

Description: A framework enabling multimodal AI models to operate a computer by viewing the screen and performing mouse and keyboard actions to achieve objectives, mimicking human computer interaction.

Key Features:

  • Compatible with various multimodal models
  • Screen observation and interaction through mouse/keyboard actions
  • Command-line interface for operation
  • Voice mode for spoken instructions
  • Cross-platform compatibility (Mac, Windows, Linux)

Technology Stack: Python, OpenAI API, Screen capture and input simulation

HAAS (Hierarchical Autonomous Agent Swarm)

Description: A groundbreaking initiative that leverages OpenAI's agent-based APIs to create a self-organizing, ethically governed ecosystem of AI agents. Inspired by the ACE Framework.

Key Features:

  • Supreme Oversight Board (SOB) consisting of high-level agents modeled after wise ethical archetypes
  • Multi-tiered agent hierarchy with specialized functions and expertise
  • Self-expanding capability allowing core agents to design and manage sub-agents
  • Robust ethical governance framework ensuring alignment with human values

Technology Stack: OpenAI APIs, Python-based implementation, Agent communication frameworks

CrewAI

Description: A multi-agent AI framework focused on sequential prompt methodology, allowing multiple specialized AI agents to collaborate on tasks.

Key Features:

  • Easy setup for multi-agent AI systems
  • Sequential prompt processing between agents
  • Configurable agent roles and behaviors

Technology Stack: Python, OpenAI API

Automated Design of Agentic Systems

Description: Research implementation focused on methodologies for automatically designing complex AI agent systems with appropriate architectures and behaviors.

Features:

  • Agent architecture design patterns and implementations
  • Automated system composition tools
  • Performance evaluation framework
  • Multi-agent coordination protocols

Technology Stack: Python, Various agent frameworks, evaluation metrics

xRx

Description: A framework for building AI-powered applications that interact with users across multiple modalities, containing reasoning applications built on top of the xRx framework.

Features:

  • Sample applications including simple templates and specialized use cases
  • Pizza store demo application
  • Shopify integration
  • Wolfram Assistant for math and physics
  • Patient information collection app for healthcare

Technology Stack: Docker, Web technologies, Various AI integrations

Ai-Dialer

Description: An implemented AI-powered system for managing phone calls with voice assistant features, transcription, and call routing capabilities.

Features:

  • Automated call initiation and management
  • Voice interaction during phone calls
  • Call transcription and summarization
  • Contact management integration

Technology Stack: Python, Telephony APIs, Speech recognition libraries

Slide-Generation

Description: Automated presentation slide generation system using AI to create educational or business content from textual inputs.

Features:

  • Content organization and structuring algorithms
  • Visual design templates and generation
  • Slide layout optimization
  • Text-to-presentation conversion pipeline

Technology Stack: Python, Presentation generation libraries, LLM integration

weekly_arxiv

Description: A project for tracking or summarizing new research papers from arXiv, focused on AI and machine learning publications.

Features:

  • arXiv API integration
  • Paper categorization
  • Summary generation
  • Research trend analysis

Technology Stack: Python, arXiv API, NLP for summarization

Pringest

Description: A tool that turns any GitHub pull request into a text diff for LLM ingestion, making it easier to feed PR changes into large language models.

Key Features:

  • Complete PR diffs conversion to unified text format
  • PR summaries and information about files changed
  • Fast generation powered by GitHub API
  • Web interface for easy access

Technology Stack: Frontend: Jinja, Tailwind CSS; Backend: FastAPI, Python; Deployment: Railway

View Resume View Software & AI Portfolio View Undergraduate Experience