DeepSeek R1: Revolutionizing AI on Azure AI Foundry and GitHub

Wednesday, January 29, 2025




DeepSeek R1 is now available on Azure AI Foundry and GitHub, marking a significant milestone in AI development. This advanced reasoning model offers powerful capabilities with minimal infrastructure investment, making cutting-edge AI more accessible to developers and enterprises.



Key Features of DeepSeek R1

Advanced Reasoning


DeepSeek R1 excels in complex reasoning tasks, making it ideal for applications requiring sophisticated problem-solving abilities.


Scalability


Built on the trusted and scalable Azure AI Foundry, DeepSeek R1 seamlessly integrates into enterprise workflows and cloud-based solutions.


Cost-Efficiency


With minimal infrastructure investment, DeepSeek R1 democratizes access to AI capabilities, making it feasible for startups and large enterprises alike.


Security and Compliance


DeepSeek R1 has undergone rigorous red teaming and safety evaluations, ensuring adherence to responsible AI principles and industry security standards.

Quick Tutorials: Getting Started with DeepSeek R1


1 Deploy DeepSeek R1 on Azure AI Foundry


Step 1: Sign in to Azure AI Foundry and navigate to the Model Catalog.


Step 2: Search for DeepSeek R1 and select the desired model variant.

Step 3: Click Deploy, configure resources (CPU/GPU), and integrate with your application via Azure OpenAI API.


Use the Azure SDK for Python to interact with the model:

import openai


client = openai.AzureOpenAI(

    api_key="YOUR_AZURE_API_KEY",

    endpoint="YOUR_AZURE_ENDPOINT"

)


response = client.Completions.create(

    model="deepseek-r1",

    prompt="What is the future of AI?",

    max_tokens=100

)


print(response.choices[0].text)



 Resources & Further Reading

Azure AI Foundry: DeepSeek R1

DeepSeek R1 GitHub Repository

Azure AI Model Deployment Guide


DeepSeek R1 brings the power of advanced reasoning AI to businesses and developers, enabling more intelligent, efficient, and scalable applications. Ready to explore? 


Free Exam, Get Ready to Fast-Track Your Career with the Microsoft Certified: Fabric Analytics Engineer Associate Certification!

Tuesday, November 5, 2024



For a limited time, the Microsoft Fabric Community team is offering 5,000 free DP-600 exam vouchers to eligible Fabric Community members.

We'll be sharing more information in this article throughout the month of November. Subscribe to stay up to date!

:loudspeaker: November 5th Update

You can request your free voucher starting on November 19th at 9:00 AM PT (Seattle, USA timezone). The URL to request your free voucher will be https://aka.ms/iamready/dp600.

Eligibility Criteria:

To be eligible for this limited-time offer, you must:

  1. Join the Fabric Community if you haven't already.

  2. Not already be a Microsoft Certified: Fabric Analytics Engineer Associate (DP-600).

  3. Register for and complete all modules in the Microsoft Learn Challenge | Ignite Edition: Fabric Challenge.

    • Pre-registration is open now!

    • On November 19th at 8:00 AM PT, you will be able to see the collection of Learn modules you must complete.

  4. Do not submit your request form before completing the challenge or your request will be denied.

  5. Be confident that you can take and pass exam DP-600 by December 31, 2024.

  6. Agree to these terms and conditions.

Already registered for the challenge? Start preparing for the exam and complete a few of the required modules now.

Steps to Prepare:

  • Watch the Get Certified! Fabric Analytics Engineer (DP-600) on-demand series.

  • Complete these Fabric learning modules.

  • Start studying for the exam.

About Microsoft Fabric:

Microsoft Fabric is an end-to-end analytics and data platform designed for enterprises that require a unified solution. It encompasses data movement, processing, ingestion, transformation, real-time event routing, and report building. Fabric integrates components like Data Engineering, Data Factory, Data Science, Real-Time Analytics, Data Warehouse, and Databases into a cohesive stack. It simplifies data integration, governance, and security across clouds and analytics engines, helping data teams collaborate, discover, and act on data with AI tools.

About the DP-600 Exam:

The DP-600 exam, titled "Implementing Analytics Solutions Using Microsoft Fabric," assesses a candidate's ability to plan, implement, and manage data analytics solutions. The exam lasts 100 minutes and includes 40-60 multiple-choice and multiple-response questions. To pass, candidates must score at least 700 out of 1000. The exam covers topics such as data modeling, data transformation, Git-based source control, SQL, DAX, and PySpark.

Free Consultation and Mentorship

Are you preparing for the DP-600 exam and looking for guidance? Reach out to Usama Wahab Khan, a Microsoft MVP and experienced technology executive, for free consultation and mentorship. Connect with him on LinkedIn or follow him on X.

Generative AI: A Primer for Users and Business Leaders

Wednesday, October 23, 2024


In the rapidly evolving world of technology, Generative Artificial Intelligence (AI) has emerged as a groundbreaking force, transforming how we create, innovate, and conduct business. This article aims to demystify Generative AI for both the novice user and the seasoned business leader, providing a detailed example to illustrate its potential.


### What is Generative AI?


Generative AI refers to the subset of artificial intelligence focused on creating new content, whether that be text, images, or even code. It leverages complex algorithms and neural networks to analyze vast amounts of data, learning patterns and styles to generate original outputs. This technology powers a range of applications, from chatbots and virtual assistants to advanced design systems.


### For the Basic User


If you're new to Generative AI, think of it as a highly advanced assistant that can help you with a variety of tasks. For instance, if you're writing an email or a report, Generative AI can suggest complete sentences or paragraphs that sound as if you wrote them yourself. It can also create realistic images or music based on your descriptions or help you code by providing snippets that fit your project's needs.


### For the Business Leader


For business leaders, Generative AI is a game-changer. It can significantly reduce the time and cost associated with content creation and product development. In marketing, for example, AI can generate personalized content that resonates with different segments of your audience, increasing engagement and conversion rates. In product design, it can rapidly prototype new ideas, speeding up the innovation cycle and bringing products to market faster.


### A Detailed Example


Imagine a retail company looking to design a new line of clothing. Traditionally, this process would involve designers sketching ideas, creating prototypes, and going through several iterations before finalizing a design. With Generative AI, the company can input current fashion trends, desired styles, and materials into an AI system, which then generates a range of design options. These options can be refined and altered until the perfect design is achieved, all within a fraction of the time it would normally take.


### Conclusion


Generative AI is not just a futuristic concept; it's a present-day tool that offers immense benefits for individuals and businesses alike. By automating and enhancing creative processes, it allows for greater efficiency, innovation, and personalization. As this technology continues to advance, it will undoubtedly open up new horizons for human creativity and enterprise.


Whether you're a basic user curious about AI's capabilities or a business leader seeking to leverage AI for competitive advantage, the journey into Generative AI is well worth embarking on. It promises to be a key driver of progress in the digital age, reshaping our approach to creation and problem-solving.

Retirement of Real-Time Streaming in Power BI: What You Need to Know and How to Migrate

Wednesday, October 16, 2024

What’s Changing?



Microsoft is making key changes to the real-time streaming capabilities in Power BI. If you’re currently using real-time semantic models for your streaming data insights, it’s essential to plan for the upcoming changes:

Starting October 31, 2024: Creation of new real-time semantic models will no longer be supported. This includes Push semantic models, Streaming semantic models, PubNub streaming, and Streaming data tiles.

By October 31, 2027: Existing real-time semantic models will be fully retired and no longer supported.

These dates are critical for organizations relying on Power BI for real-time insights. Microsoft has committed to working with existing customers on migration strategies leading up to the 2027 deadline, with potential for date adjustments as necessary.

What’s the Alternative?



Microsoft recommends transitioning to Real-Time Intelligence (RTI) solutions available in Microsoft Fabric, which provides a more comprehensive platform for real-time insights and data streaming. Fabric’s capabilities go beyond what Power BI’s real-time streaming offered, delivering robust solutions for event-driven scenarios, data logs, and streaming data.

For new real-time streaming requirements, leveraging Microsoft Fabric is the best way forward. It enables data streaming from multiple sources, geospatial analysis, and actionable insights all within a unified platform.

Key Features of Microsoft Fabric Real-Time Intelligence



1. Centralized Real-Time Hub: Fabric’s Real-Time Hub serves as the central repository for streaming data, allowing easy access, exploration, and sharing across your organization. It integrates with sources like Azure Event Hubs, Azure IoT Hub, and more, ensuring seamless data flow.

2. Event Streams: With a no-code interface, you can capture and transform real-time events from various sources, including Azure, AWS, and Google, and route them to the desired destinations.

3. Event Processing: Fabric allows for real-time data cleansing, filtering, transformation, and aggregation. You can also create derived streams for more tailored data sharing and processing.

4. Eventhouses: These specialized engines are designed for time-based event analytics, enabling quick and powerful querying of both structured and unstructured data.

5. Visualization & Insights: Seamlessly integrate with Power BI to visualize your data and create dashboards and reports for real-time insights. Alerts can trigger actions based on changing data patterns, turning insights into actions.

6. Data Activator: Fabric’s Data Activator lets you respond to real-time data by triggering alerts and actions, such as sending notifications or invoking workflows when certain data conditions are met.


Moving forward  to Migrate Your Existing Real-Time Models

If you’re using real-time semantic models in Power BI, the transition to Microsoft Fabric should be part of your future planning. Key steps in the migration process include:

Review Current Models: Evaluate the existing real-time semantic models in use and assess their role in your business workflows.

Explore Fabric’s Capabilities: Understand how Fabric’s Real-Time Hub, Event Streams, and Eventhouses can replace or enhance your current real-time streaming setup.

Plan Your Migration: Begin planning to transition before the 2027 deadline. Work closely with Microsoft or a certified partner to ensure a smooth migration.


For further guidance, visit aka.ms/RTIblog, which will be continually updated with migration resources and best practices.


Final Thoughts


While the retirement of real-time streaming in Power BI marks the end of an era, it also opens the door to more powerful and flexible real-time intelligence solutions in Microsoft Fabric. By preparing now and exploring the possibilities in Fabric, you can continue to harness the power of real-time data to drive smart, timely decisions across your organization.


Don’t wait until the last minute—start your migration planning today to ensure you stay ahead of these changes!

Introducing Meta LLaMA 3: A Leap Forward in Large Language Models

Thursday, April 18, 2024




Meta has recently unveiled its latest innovation in the realm of artificial intelligence: the LLaMA 3 large language model. This state-of-the-art model represents a significant advancement in AI technology, offering unprecedented capabilities and accessibility.



What is LLaMA 3?


LLaMA 3 is the third iteration of Meta's large language model series. It is an open-source model that has been fine-tuned with instructions to optimize its performance across a wide array of tasks. The model comes in two sizes: one with 8 billion parameters and another with a colossal 70 billion parameters.

Features and Capabilities



The LLaMA 3 models are designed to excel in language understanding and generation, making them highly effective for applications such as dialogue systems, content creation, and complex problem-solving. Some of the key features include:


-Enhanced Reasoning

LLaMA 3 demonstrates improved reasoning abilities, allowing it to handle multi-step problems with ease.

-Multilingual and Multimodal Future

 Plans are underway to make LLaMA 3 multilingual and multimodal, further expanding its versatility.

Extended Context Windows

 The new models support longer context windows, enabling them to maintain coherence over larger text spans.


The Meta Llama 3 models have been enhanced with a substantial increase in training tokens, reaching 15trillion, which greatly improves their ability to grasp the nuances of language. The context window has been expanded to 8,000 tokens, effectively doubling the previous model's capacity and allowing for the processing of more extensive text excerpts, which aids in making more informed decisions. Additionally, these models employ a novel Tiktoken-based tokenizer that boasts a128,000-token vocabulary, resulting in a more efficient encoding of characters per token. Meta has observed improved performance in both English and multilingual benchmark assessments, confirming the models' strong capabilities in handling multiple languages.


Unmatched Performance Excellence


The introduction of our 8B and 70B parameter LLaMA 3 models marks a significant advancement beyond the capabilities of LLaMA 2, setting a new benchmark for large language models (LLMs) at these scales. Enhanced pretraining and refined post-training techniques have elevated our models to the pinnacle of performance, making them the premier choice in the current landscape for 8B and 70B parameter models. Notable enhancements in our post-training processes have led to a considerable decrease in incorrect rejections, bolstered model alignment, and enriched the variety of responses generated by the models. Furthermore, we've observed a remarkable enhancement in functions such as logical reasoning, code creation, and adherence to instructions, rendering LLaMA 3 more adaptable and responsive to user guidance.

Accessibility and Community Support



In line with Meta's commitment to open innovation, LLaMA 3 is made available to the broader community. It can be accessed on various platforms, including AWS, Databricks, Google Cloud, and Microsoft Azure, among others¹. This move is intended to foster a wave of AI innovation across different sectors.


It's now available on Azure 

https://techcommunity.microsoft.com/t5/ai-machine-learning-blog/introducing-meta-llama-3-models-on-azure-ai-model-catalog/ba-p/4117144


Trust and Safety


Meta has introduced new trust and safety tools, such as LLaMA Guard 2 and Code Shield, to ensure the responsible use of LLaMA 3. These tools are part of a comprehensive approach to address the ethical considerations associated with deploying large language models¹.


The Impact of LLaMA 3


The release of LLaMA 3 is poised to have a profound impact on the AI landscape. By providing a powerful tool that is openly accessible, Meta is enabling developers and researchers to push the boundaries of what's possible with AI. The model's capabilities in understanding and generating human-like text will unlock new possibilities in various fields, from education to customer service.


As we look to the future, LLaMA 3 stands as a testament to Meta's dedication to advancing AI technology while maintaining a focus on ethical and responsible development. It's an exciting time for AI, and LLaMA 3 is at the forefront of this technological revolution.

More details 

(1) Introducing Meta Llama 3: The most capable openly available LLM to date. https://ai.meta.com/blog/meta-llama-3/.

(2) Meta Llama 3. https://llama.meta.com/llama3/.


#Meta #llama #Azure #MVPBuzz #generativeai #GenAI #LLM #Opensource 


Power BI With Copilot

Sunday, April 14, 2024




Get Started with Copilot for Power BI and Create Reports Faster

Ready to unlock the power of AI in your data analysis? Copilot, a new feature in Power BI Desktop, is here to help you create reports faster and easier. With Copilot's assistance, you can generate report summaries, suggest report content, and even create entire report pages based on your high-level instructions.

What You'll Need to Use Copilot:

  • Access: You'll need write access to a workspace assigned to a paid Power BI capacity (P1 or higher) or a paid Fabric capacity (F64 or higher) with Copilot enabled by your administrator.
  • Power BI Desktop: Ensure you're using the latest version of Power BI Desktop.

Getting Started with Copilot:

  1. Enable Copilot (Admin): Your administrator needs to enable Copilot in Microsoft Fabric and activate the tenant switch.
  2. Open the Copilot Pane: Click the Copilot icon in the ribbon to open the Copilot pane.
  3. Welcome and Workspace Selection: The first time you use Copilot, a dialog will appear prompting you to choose a compatible workspace. Select any workspace assigned to the required capacity.
  4. Start Your Interaction: Once you've selected a workspace, you'll see a welcome card. Click "Get started" to begin using Copilot.

How Copilot Can Help You:

  • Summarize Your Data Model: Gain a clearer understanding of your data with Copilot's summaries of your Power BI semantic model. This can help you identify key insights and streamline your report building process.


  • Suggest Report Content: Stuck on what to include in your report? Copilot can analyze your data and propose relevant topics for you to explore. From the Copilot pane, select "Suggest content for a report" to get started.
  • Create Report Pages: Save time crafting reports from scratch. Provide Copilot with a high-level prompt related to your data, and Copilot will generate a customizable report page with relevant tables, fields, measures, and charts. Here are some examples of prompts you can use:
    • Analyze performance across different shifts based on metrics like good count, reject count, and alarm count.
    • Evaluate production line efficiency and overall equipment effectiveness.
    • Compare costs, materials, and their impact on production for each product


Important Considerations:

  • The Copilot button will always be visible in the ribbon, but functionality requires signing in, admin enabled tenant settings, and workspace access as mentioned earlier.
  • The workspace you select for Copilot usage doesn't have to be the same one where you publish your report.
  • Copilot is currently in preview, and its responses are generated using AI, so always double-check your work for accuracy.
  • There are limitations for creating report pages with certain connection modes like live connect to SSAS and AAS, and real-time streaming in Power BI.

Stay Updated:

Keep an eye out for the latest Copilot enhancements by following the monthly Power BI feature summary blogs.


https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-introduction


Embrace AI-powered report creation with Copilot and transform your data analysis workflow!

RAG live Demo with Azure and openAi

Saturday, April 6, 2024