Local corner · September 2026
Each month, five people whose public work helps others run AI models on their own hardware: local runtimes, on-device frameworks, efficient fine-tuning, and open-weight models. The lineup changes at the start of each month.
Works on Ollama, an open-source tool for running open models on your own machine
Jeffrey Morgan works on Ollama, an open-source tool for running open models on your own machine; his GitHub profile says he is building Ollama. He writes code for the project: the repository's commit history shows him authoring recent changes to context-window handling and to model support in its MLX engine.123
Works on Transformers.js at Hugging Face, a library for running Transformers models in the browser
Joshua Lochner works on Transformers.js at Hugging Face, a library that runs Transformers models directly in the browser with no server required. He is a byline author of Hugging Face's release announcements for Transformers.js v3 in 2024, which added WebGPU support, and v4 in 2026. His GitHub profile describes his work as bringing machine learning to the web.1234
Head of Post-Training at Liquid AI; author of the LLM Course and its quantization tutorials
Maxime Labonne is Head of Post-Training at Liquid AI, where his work is post-training the company's LFM2 and LFM2.5 models. He is listed among the contributors to the LFM2 technical report, whose models ship with open weights and deployment packages for tools such as llama.cpp and ExecuTorch. He also publishes the LLM Course on GitHub, which includes his tutorials on quantizing models with GGUF and llama.cpp, GPTQ, and ExLlamaV2, and an AutoQuant notebook for producing quantized versions of a model.143
Developer Experience Lead at Google DeepMind, working on Gemma launches and developer guides
Omar Sanseviero leads developer experience at Google DeepMind for AI Studio, the Gemini API, and Gemma, including the developer side of model launches. He describes himself as closely involved in Google DeepMind's open model efforts, among them MedGemma and EmbeddingGemma. He has co-authored Google Developers Blog guides to Gemma 3n, a mobile-first model for on-device use, and Gemma 4 12B, which the guide says is small enough to run locally on laptops with a dedicated GPU. He was previously Chief Llama Officer and Head of Platform and Community at Hugging Face.123
Executive Director of EleutherAI; first author of the Pythia open model suite
Stella Biderman is the Executive Director of EleutherAI. She is first author of the Pythia paper, which introduced 16 language models from 70M to 12B parameters, all trained on public data in the same order, with 154 checkpoints released for each model so researchers can study how models change over training and scale. She is also among the credited authors of the Language Model Evaluation Harness, EleutherAI's open framework for evaluating language models.1234
Profiles use only public, professional information — roles, projects, and pages people publish themselves — with a source for each statement. They never include locations, nationality, ages, photos, or personal details. Everyone featured is connected to an organization or project in this catalog. Selection is editorial, is not a ranking, and is never influenced by tips.
To suggest someone, correct a profile, or ask to be removed, email usasihq@gmail.com. Removal requests are honored.
Help keep USASI useful.
Optional. No USASI account required. Payment takes place on the linked provider’s website (Buy Me a Coffee).
About supporting this project