Custom AI Model in AI Foundry

David Yoon 0 평판 포인트 Microsoft Regional Director
2025-11-11T00:02:40.4366667+00:00

Hello,

I have few questions about Azure AI Foundry like as below,

  1. Can users utilize custom generative AI models created by themselves in AI Foundry? Is there functionality to directly upload models created independently, beyond fine-tuning existing Foundry models? Currently, I understand this feature isn't available, and models must be registered as community models in the marketplace for use.
  2. How can I use an AI model created in Azure Machine Learning directly in AI Foundry? I've seen guides stating it requires API calls, but I'm curious about the actual implementation instructions.

_
Thanks in advance_

Foundry Tools
Foundry Tools

이전 명칭은 Azure AI Services 또는 Azure Cognitive Services였으며, Microsoft Foundry 플랫폼 내의 사전 구축된 AI 기능을 통합한 컬렉션입니다

댓글 0개

답변 2개

정렬 기준: 최신순
가장 유용함 최신순 등록순
  1. David Yoon 0 평판 포인트 Microsoft Regional Director
    2025-11-11T10:15:34.3966667+00:00

    Hi, Anshika

    Thanks for your answer but I want to know how to add custom AI model made by myself instead find tunned foundation model.

    I'm not sure if we can add custom AI model in Azure AI Foundry except Marketplace.

    Hoep to see your answer and insight as well.

    Regards,
    David

    이 대답이 도움이 되었나요?

  2. Anshika Varshney 15,625 평판 포인트 Microsoft 외부 직원 중재자
    2025-11-11T01:38:02.3966667+00:00

    Hello David Yoon ,

    thanks for your question! Azure AI Foundry supports creating and customizing models through fine-tuning rather than training from scratch. Here’s how you can approach it:

    Key Steps to Create a Custom Model

    1. Prepare Your Data
      • Collect a clean dataset relevant to your use case (classification, summarisation, domain-specific tasks).
      • Split into training and validation sets in JSONL format for text-based models.
    2. Choose Fine-Tuning Method
      • In Azure AI Foundry portal, navigate to Model Catalog → Azure OpenAI or other providers.
      • Select a base model (e.g., GPT‑4.1 or GPT‑4.1‑mini for smaller tasks).
      • Use LoRA (Low-Rank Adaptation) for efficient fine-tuning—this reduces cost and speeds up training.
    3. Run Fine-Tuning Job
      • Options: Portal, Python SDK, or REST API.
      • Monitor job status in the portal and review logs for any errors.
    4. Deploy Your Custom Model
      • After fine-tuning completes, deploy via Azure AI Foundry Models.
      • Choose serverless API or managed compute based on your performance and cost needs.
    5. Test and Iterate
      • Use the Playground in Foundry to validate outputs.
      • Adjust training data or parameters if results don’t meet expectations.

    Benefits of Fine-Tuning

    1. Improves accuracy for domain-specific tasks.
    2. Reduces token usage and latency compared to prompt engineering.
    3. Enables compliance with organizational tone and policies

    If you need advanced control (like embeddings or multi-agent orchestration), you can integrate Azure AI Foundry Agent Service or use Azure Machine Learning pipelines for complex workflows.

    References:
    Fine-tune models with Azure AI Foundry and Customize a model in AI Foundry.
    I hope this clears things up! If you need any further assistance or run into any issues, please don’t hesitate to reach out I’ll be happy to help.

    Thankyou!

    이 대답이 도움이 되었나요?

    댓글 0개

답변

질문 작성자는 답변을 '승인됨'으로 표시하고, 중재자는 답변을 '추천됨'으로 표시할 수 있습니다. 이를 통해 사용자는 해당 답변이 작성자의 문제를 해결했다는 것을 알 수 있습니다.