4 pointsby thangngoc892 days ago1 comment
  • Jugurtha2 days ago
    Hi. First of all, congratulations! This is nice to hear.

    I ran a consulting company that did bespoke ML products then built a product that supported us doing that (MLOps) to ship faster, then turned it into a product company with that product (it offered real-time collaborative notebooks, training, tracking, packaging, deploying, and monitoring models on all major cloud providers and on customer data sources).

    I've written a Tweet that could be helpful: https://x.com/jugurthahadjar/status/1310668293305499653

    It depends on the "maturity" of your clients. Some have technical capabilities and they'll be just fine with your model: they can deploy and monitor and use it within their systems. They may or may not have monitoring and you'll probably have to deal with drift as time goes by.

    Others will want to interact with some interface (REST API or something).

    Some will want an entire application specifically for a role/department: front-end, user management, etc.

    And some will just want to use it and you deal with everything. Have you thought of making a product for that niche, so it's not just the model and the pipeline, but a whole application that solves this niche's problem?

    • thangngoc892 days ago
      Hi. Thank you so much for sharing your experience. As I mentioned, I’m quite new to this, and your answer brought up some important questions I hadn’t considered. I believe it would be best to discuss these further with the clients.

      Would it be alright if I reached out to you via email in the future if I have any additional questions? I noticed your email in your profile.