{"slug": "local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted", "title": "Local LLMs/VLMs for Physical AI engineering workflows — model suggestions wanted", "summary": "Afrid Ali is developing EdgeAI Forge, an early-stage local-first architecture for using specialized AI agents in industrial vision, ROS 2, robotics, and edge-deployment workflows, and is seeking model recommendations and evaluation ideas for local coding agents, VLMs for industrial inspection, smaller reasoning models, embedding models, and runtimes across workstation GPUs and Jetson-class devices. The proof of concept routes Planner, Vision, and ROS prompts to Ollama, with optional cloud escalation and explicit human review before physical deployment.", "body_md": "**I am experimenting with **EdgeAI Forge**, an early-stage local-first architecture for using specialized AI agents in industrial vision, ROS 2, robotics, and edge-deployment workflows.**\n\n**The intended split is:**\n\n**- local models for frequent coding, planning, documentation, private-data processing, and visual reasoning;**\n\n**- optional cloud escalation for difficult or large-context tasks; and**\n\n**- explicit human review before any physical deployment or machine-control action.**\n\n**The current proof of concept routes Planner, Vision, and ROS prompts to Ollama. I am looking for practical model recommendations and evaluation ideas for:**\n\n**- local coding agents that produce structured Python/C++/ROS 2 artifacts;**\n\n**- VLMs for industrial image inspection and pipeline selection;**\n\n**- smaller reasoning models for planning and validation;**\n\n**- embedding models for project memory and retrieval; and**\n\n**- runtimes that work well across workstation GPUs and Jetson-class devices.**\n\n**I am especially interested in measured tradeoffs—VRAM, latency, structured-output reliability, quantization, tool use, and license constraints—rather than leaderboard scores alone.**\n\n**Project: GitHub - afridali123/EdgeAI_Forge · GitHub**\n\n**Longer background: EdgeAI Forge: My Journey Toward a Local Agentic AI Platform for Industrial Automation**\n\n**This is an early architecture discussion, not a finished product announcement. Suggestions for models, datasets, evaluation harnesses, or related projects are welcome.**", "url": "https://wpnews.pro/news/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted", "canonical_source": "https://discuss.huggingface.co/t/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted/178524#post_1", "published_at": "2026-08-09 14:19:00+00:00", "updated_at": "2026-08-09 14:44:04.450031+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "computer-vision", "ai-agents", "ai-infrastructure"], "entities": ["Afrid Ali", "EdgeAI Forge", "Ollama", "ROS 2", "Jetson"], "alternates": {"html": "https://wpnews.pro/news/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted", "markdown": "https://wpnews.pro/news/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted.md", "text": "https://wpnews.pro/news/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted.txt", "jsonld": "https://wpnews.pro/news/local-llms-vlms-for-physical-ai-engineering-workflows-model-suggestions-wanted.jsonld"}}