{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreia7qe56rvf76gtlkk5halyp6eajlcqtaxbncfp3ijjuopqms7mjti",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mgyfi7pnbwx2"
  },
  "path": "/t/looking-for-feedback-from-pytorch-users-on-an-ml-prototyping-workflow-idea/174249#post_1",
  "publishedAt": "2026-03-13T17:33:15.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "https://aimlse.org"
  ],
  "textContent": "Hi everyone,\n\nI’ve been working on a project called **AIMLSE** and was hoping to get some feedback from people here who regularly build or experiment with **PyTorch models**.\n\nOne idea behind it is something called an **Adaptive Coding Interface (ACI)**. The concept is that you write a short **description of the ML project you want to build** , and the system recommends **publicly contributed helper functions** that might help structure the project (for example things related to training loops, dataset handling, or evaluation).\n\nThe goal is to make it easier to go from **idea → working prototype** , especially for people experimenting with different models.\n\nI’m currently looking for a few people who would be interested in **trying the platform and giving feedback** , mainly on things like:\n\n  * whether the helper function recommendations are actually useful\n\n  * whether the workflow feels natural when building a project\n\n  * what parts feel confusing or unnecessary\n\n\n\n\nIt’s currently focused on **PyTorch-based projects** , but the long-term idea is to support more workflows.\n\nIf anyone here enjoys trying new ML tools or experimenting with development workflows, I’d really appreciate your thoughts.\n\nYou can take a look here: **https://aimlse.org**\n\nThanks, and I’d love to hear what people think.",
  "title": "Looking for feedback from PyTorch users on an ML prototyping workflow idea"
}