Jay JohnsonAI, Architecture, Strategy, Adoption

Jay Johnson

CEO Matlok.ai · Building Self-Hosted AI Platforms

  • I design, build, and run AI platforms that generate enterprise ROI.
  • I lead teams on-premise and on clouds with proprietary and open-source models, RAG, agentic coding, data pipelines, training and fine-tuning models, and evangelizing AI adoption.
  • Over +20 years of building platforms, from a Fortune 25 automotive cloud with over a million connected cars to all trains in North America to custom MEG brain-imaging software, and a patent in OpenShift.

I help teams automate systems and workflows with AI and human-ai teaming strategies. I enjoy sharing how technology works on GitHub and HuggingFace: 2500+ AI collections, 19 published datasets (~1.5 TB) and 2 models on Hugging Face.

Projects

  1. OATs Agentic Coder

    Open Agent Tools coder: an agentic coding assistant with a tool belt of 23 tools plus bash.

    agentscodingpython

  2. AI - Stock Analysis Engine

    Build and tune investment algorithms with distributed workers, datasets and AI models.

    fintechaidistributed

  3. AI Network Defense

    AntiNex: deep neural networks trained to detect network attacks from captured traffic.

    securitydeep learning

  4. AI Training

    Train AI models through a REST API: Django, Swagger and JWT on top of Keras and TensorFlow.

    trainingrest api

  5. 1.5 TB Open Source Multimodal Dataset

    We share large, open source multimodal datasets for training and fine-tuning AI to write python and build AI models, and curate collections of guides, papers, datasets, models and tools like frankenmerging AI models.

    More extracted images and mp3s from the datasets: python-copilot-image-and-audio-examples.

    datasetsmultimodalhugging face

  6. Latest AI model: tinyllama-cinder-openhermes-32k

    A frankenmerge of 5 AI models into 1 AI model. Click the Compute button on its page to talk with it.

    modelmerge