Internectual Labs / The Team

The people behind
the experiments.

Independent research. Practical engineering.
A shared drive to learn, test, and build something meaningful.

01 / Research & directionOwner & Founder
Cody Dixon’s GitHub profile image

@CodeMasterCody3D

Cody Dixon

Independent AI researcher & builder

33 · Hueytown, Alabama

Background & journey

I’m Cody, owner and founder of Internectual Labs, an independent AI researcher and builder. I started by overclocking hardware and digging into Linux, learning through forums and trial and error. Getting RuneScape running smoothly on Android through Linux was an early win. Later, I pushed an i7-8086K with a compact 120 mm liquid cooler and tuned mining rigs for efficiency. Those experiments taught me to balance performance, power, and stability.

Technical focus

AI took me from not knowing how to code to building my own projects. Now I’m drawn to the same question that drove my hardware experiments: how much more can we get out of less? I want to shrink AI models while retaining their intelligence, connecting model efficiency with the practical limits of the hardware running them.

Work at the lab

That drive led to TAARDIS-27B, the lab’s full-ternary model project. My TK Integers work explores ternary weight formats and GPU kernels. IQ wave models are more experimental, exploring geometry and interacting waves in place of standard layers. I also build tools: GRIP helps AI-assisted projects keep track of where they left off, and GNU-AI-Flow is a visual workspace for model components and training pipelines.

Personal approach

I love pushing limits, questioning assumed boundaries, and improving things until the results hold up. When something fails, I want to understand why and keep working at it. I value friendly, honest conversation and hearing what people actually care about. The lab grows from that same curiosity: build something, test it, share what you learn, and keep going.

  • Ternary quantization
  • IQ wave models
  • Tools for builders

At Internectual Labs

Project Contributions

  • TK Integers Creator & Lead

    Ternary weight formats, GPU kernels, and model precision tools.

  • TAARDIS-27B Creator & Lead

    The lab’s full-ternary model project and runtime development.

  • GRIP Creator & Lead

    Project continuity and progress tracking for AI-assisted work.

  • IQ Wave Models Creator & Lead

    Experimental architectures using geometry and interacting waves.

  • GNU-AI-Flow Creator & Lead

    A visual model and training workspace based on GNU Radio Companion.

Selected public projects

Explore their work

Cody’s work on GitHub

  • AI / Ternary computing

    TK Integers

    Ternary weight packing, GPU kernel experiments, and tools for exploring model precision.

  • AI / Developer tools

    GRIP

    Keeps AI-assisted projects moving with saved context, progress, and a record of what was tested.

  • AI / Runtime fork

    TAARDIS Runtime

    A llama.cpp fork for full-ternary models, ternary KV cache experiments, and CPU / CUDA execution.

    Built on Prism ML’s llama.cpp fork.

  • Android / 3D printing

    DroidKlipp

    Turns an Android device into a KlipperScreen display, with USB ADB forwarding and Wi-Fi fallback.

  • Android / Slicer fork

    OrcaSlicer Mobile

    Brings the OrcaSlicer engine to Android with model imports, print profiles, and multi-color painting.

    Built on Slice Beam and OrcaSlicer. Experimental / alpha.

  • 3D printing / Klipper

    Filament Change Tracking

    Tracks manual filament changes and prompts the next color for multi-color prints using OrcaSlicer data.

  • 3D printing / Plugin fork

    OctoPrint BambuPrinter

    Cody’s fork of the Bambu Lab printer integration for OctoPrint, including FTP upload changes.

    Based on jneilliii’s OctoPrint-BambuPrinter.

  • AI / Dataset tools

    repo2dataset

    Collects repository source, bug-fix pull requests, and fix commits into structured JSONL training data.

  • Linux / Input tools

    RemoteMouse Wayland Bridge

    Bridges XTest-based remote input into a native KDE Plasma Wayland session through virtual input devices.

  • Developer tools / Game assets

    torch-mcp

    An MCP interface for the HarbourMasters Torch asset extraction and modding toolchain.

  • Linux / Game preservation

    Tribes 2 Docker

    Packages a compatibility environment for running the original Linux version of Tribes 2 on modern systems.

  • Audio / Research in progress

    TribesTalk

    An experimental sample-based voice engine and preparation workflow for custom Tribes 2 speech.

    In development; not a finished voice-bank release.

02 / Engineering & implementationNew Team Member
Mohammed Ain Tomar

@Mohammed18-19

Mohammed Ain Tomar

Junior AI Engineer | Python Backend | LLM & Agentic AI

22 · Kénitra, Morocco

Background & journey

I’m Mohammed, a self-taught developer learning since 2023. I originally wanted to study computer science and am now studying Electronics and Information Processing Engineering. Alongside that, I built my Python and software skills, then moved into machine learning, LLMs, RAG, and agents. I’m early in my professional career, with no company employment yet; my experience comes from building projects and contributing to open source.

Technical focus

I’m interested in reliable LLM applications: tool calling, memory, agent orchestration, and source-grounded RAG with hybrid retrieval and embeddings. My backend work uses Python REST APIs, Flask, PostgreSQL, pgvector, and SQLAlchemy. I’m strengthening my ML fundamentals and model evaluation skills, with testing, debugging, Docker, CI/CD, and Git as part of how I build.

Projects & planned lab work

My projects include AgriMarket AI, combining FAOSTAT production and NASA POWER climate data for Moroccan agricultural intelligence, backend pipelines, forecasting, and LangGraph agents; and CodeAtlas, a codebase intelligence tool using hybrid retrieval and source-grounded answers. My open-source work includes embedded bootloader security, regression tests, and computer vision inference validation.

At Internectual, I want to help test TK Integers, evaluate Qwen quantization, and investigate training stages, checkpoint restoration, reliability, and stage integration. I also want to explore multi-agent reasoning and systematic testing while learning more about quantization, training, distillation, and reasoning optimization. These are intended areas of contribution as I get started.

Personal approach

I learn by implementing ideas, understanding failures, and improving what I build. I care about helping developers, sharing knowledge, and collaborating in open source. I want my work to make a meaningful contribution beyond a line on my CV.

  • LLM & agent systems
  • Python backends & RAG
  • Testing & evaluation

At Internectual Labs

Project Contributions

    New to the lab. Contributions will appear here as work is completed.

    Selected public projects

    Explore their work

    Mohammed’s work on GitHub

    • AI / Code intelligence

      CodeAtlas

      A codebase research assistant using retrieval and source-linked answers to help developers understand repositories.

    • AI / Data analysis

      AgriMarket Intelligence

      An agricultural analytics project combining data pipelines, forecasts, and multi-agent workflows.

    • Web / Interactive design

      Portfolio

      A responsive personal portfolio built with Next.js, animated interfaces, and Three.js visuals.

    • Open source / Contribution

      eBoot

      Contributed shared image-slot boundary validation across boot, recovery, and application-loading paths.

      Contribution to embeddedos-org/eBoot; not the original project author.

      View merged contribution

    How we work

    Curiosity, with follow-through.

    Explore an idea. Build an experiment. Test what holds up. We bring research and engineering together through open collaboration, shared knowledge, and results others can examine.

    Explore the lab’s projects