DEV Community: Wriju's Blog

Demystifying Microsoft Foundry: The Ultimate Hub for AI & Agents

Wriju's Blog — Tue, 10 Feb 2026 18:30:00 +0000

Are you looking to build the next generation of intelligent applications but feeling overwhelmed by the complexity of infrastructure and model management?

In our latest video on TechTalks - Wriju, we dive deep into Microsoft Foundry (formerly Azure AI Studio). Whether you are an AI engineer or a cloud architect, understanding how this unified platform orchestrates models, data, and agents is key to scaling your AI projects.

📝 Summary

In this whiteboard session, Wriju provides a comprehensive overview of Microsoft Foundry. The video breaks down the platform into its core components:

Model Catalog: Access to over 100+ LLMs (including OpenAI, Microsoft, and other third-party providers) without managing the underlying infrastructure.
Foundry Agent Service: A Platform-as-a-Service (PaaS) offering that lets you build and host semi-autonomous agents that can interact with data and execute logic.
Data Integration (RAG): Seamlessly connecting your AI to various data sources for Retrieval-Augmented Generation.
Control Plane: A centralized hub for monitoring, security (Microsoft Entra ID), and governance.
Developer Ecosystem: How Foundry integrates with tools you already use, like VS Code, GitHub Copilot, and the Foundry SDK.

🚀 Why You Should Watch It

Architecture First: Instead of just looking at code, you’ll see a visual "big picture" diagram of how components stitch together [08:30].
Infrastructure-Free AI: Learn how to deploy models via simple drop-down options without worrying about networking or VM management [01:02].
Future-Ready SDKs: Discover the new Agent SDK, which brings together the best of Semantic Kernel and AutoGen [07:27].
Edge & Local Scenarios: Understand how Foundry Local allows you to build and test even in low-internet or on-prem environments [06:51].

📺 Watch the Video

Ready to jump in? Watch the full overview here:
https://youtu.be/1FoS8C_0sIY

📚 Official Microsoft Documentation

To further your learning, explore these official Microsoft resources:

Microsoft Foundry Overview – Learn the basics of the unified AI platform.
Foundry Agent Service Documentation – A deep dive into building and orchestrating AI agents.
Get Started with the Foundry SDK – Technical guide on using the Python and .NET SDKs for your projects.
Microsoft Foundry Models – Explore the 11,000+ models available in the catalog.

🔔 Stay Updated

If you found this video helpful, don't forget to subscribe to the channel for more deep dives into Azure, AI, and modern application development!

👉 Subscribe to TechTalks - Wriju

Azure Kubernetes Best Practices

Wriju's Blog — Thu, 15 Aug 2024 08:36:40 +0000

Azure Kubernetes Service (AKS), following best practices can help ensure your clusters are secure, efficient, and reliable. Here are some key areas to focus on:

Cluster and Pod Security

Secure API Server Access: Use Microsoft Entra ID and Kubernetes RBAC to control access to the API server.
Pod Security Policies: Implement policies to restrict pod capabilities and enforce security standards.
Container Image Security: Use trusted registries and scan images for vulnerabilities.

Networking and Connectivity

Network Policies: Define network policies to control traffic between pods.
Ingress Controllers: Use ingress controllers and web application firewalls (WAF) to manage and secure external access.

Resource Management

Resource Quotas: Set resource quotas to manage resource allocation and prevent overuse.
Pod Disruption Budgets: Use these to maintain application availability during maintenance.

Monitoring and Logging

-Centralized Logging: Implement centralized logging for better visibility and troubleshooting.
-Monitoring Tools: Use tools like Azure Monitor and Prometheus to track cluster performance and health.

Business Continuity and Disaster Recovery

Regular Backups: Schedule regular backups of your data and configurations.
Disaster Recovery Plans: Develop and test disaster recovery plans to ensure quick recovery from failures.

Cost Optimization

Right-Sizing Resources: Choose appropriate VM sizes and scale resources based on demand.
Auto-Scaling: Implement auto-scaling to adjust resources dynamically.

Personal Economic Resilience

Wriju's Blog — Sat, 20 Jan 2024 06:10:01 +0000

While older generations may have prioritized "living within your means," today's credit-driven culture and the expansion of virtual money can create a sense of abundance that masks underlying limitations. This inflated economy, where banks lend more than they hold, is inherently fragile and prone to sudden crashes, as we've seen with companies and even entire countries buckling under debt burdens. So, what can we do as individuals to navigate this uncertain landscape?

The key lies in making conscious choices and building financial resilience. This doesn't necessarily mean rejecting all credit or living like a hermit. Instead, it's about:

Understanding our relationship with money: Examining our spending habits, values, and goals to develop a healthy financial mindset.
Prioritizing essential spending: Avoiding impulsive purchases and focusing on basic needs like housing, food, and healthcare.
Building an emergency fund: Having a safety net to cushion unexpected expenses and avoid getting trapped in debt cycles.
Planning for the future: Investing in retirement savings and other long-term goals, even if it starts with small, consistent contributions.
Borrowing responsibly: Understanding the terms and risks of credit, only borrowing for essential needs with clear repayment plans.
Supporting sustainable businesses: Choosing companies that prioritize responsible financial practices and long-term value creation.

By taking these steps, we can become financially empowered individuals, less susceptible to the whims of an inflated economy. We can contribute to a healthier financial ecosystem where responsible choices, not excessive debt, drive growth and create a more secure future for ourselves and future generations.

Remember, financial resilience is a journey, not a destination. It's about making informed choices, building knowledge, and adapting to changing circumstances. By taking control of your finances, you can weather any storm and build a stronger foundation for your future.

Linux is Important

Wriju's Blog — Fri, 19 Jan 2024 17:04:48 +0000

Linux is a powerful and versatile operating system that powers many modern technologies, such as Android devices, servers, and supercomputers. Learning Linux can boost your resume and career prospects, especially if you are interested in system administration, software development, or cybersecurity. Some of the reasons why an information technology engineer should learn Linux are:

Linux is open source, which means you can access and modify its source code, offering greater flexibility and customization options.
Linux has a large and active community of developers and enthusiasts who provide free support and resources for learning and troubleshooting.
Linux has high security and stability, as it is less prone to viruses and malware, and can run on various hardware platforms without crashing.
Linux offers a wide range of applications and tools for various purposes, such as web development, DevOps, virtualization, and containerization.
Linux has a command-line interface that gives you powerful control over the system and its functions, such as file management, data manipulation, and scripting.

To learn Linux, you should start with the basics of the Linux file system and the command line. Here are some of the main commands you should learn:

ls (list files)
pwd (get current directory)
cd (change directory)
mkdir (create directory)
touch (create file)
nano (update file)
cp (copy file or directory)
mv (move or rename file or directory)
rm (remove file or directory)
cat (display file content)
grep (search for a pattern in a file)
find (search for files or directories)
ps (show processes)
kill (terminate a process)
ping (test network connectivity)
ssh (secure shell connection)
scp (secure copy files)
curl (transfer data from or to a server)
tar (create or extract compressed files)

These are some of the essential Linux skills that every information technology engineer should learn. However, there is much more to explore and master in Linux, depending on your goals and interests. You can find many online courses, tutorials, books, and forums that can help you learn Linux at your own pace and level.

How Token works in Open AI

Wriju's Blog — Fri, 19 Jan 2024 16:43:14 +0000

A token in OpenAI is a unit of text that represents a common sequence of characters. For example, the word "hamburger" gets broken up into the tokens "ham", "bur" and "ger", while a short and common word like "pear" is a single token¹. Tokens are used by OpenAI's large language models to process and generate text.

The number of tokens in a piece of text is calculated by applying a tokenizer, which is a function that splits the text into tokens. Different models use different tokenizers, so the same text may have different token counts depending on the model. For example, newer models like GPT-3.5 and GPT-4 use a different tokenizer than previous models and will produce different tokens for the same input text.

When a prompt is more than the limited token, the model will truncate the prompt to fit within the limit. For example, if the limit is 2048 tokens and the prompt is 2500 tokens, the model will only use the first 2048 tokens as input and ignore the rest. The output is also limited by the token count, and the model will stop generating text when it reaches the limit. For example, if the output limit is 512 tokens and the model has generated 500 tokens, it will only generate 12 more tokens before stopping.

Hope this helps.

Token in OpenAI Platform. https://platform.openai.com/tokenizer.

Service Level Agreement (SLA) and Composite SLA

Wriju's Blog — Fri, 19 Jan 2024 16:32:20 +0000

A service level agreement (SLA) is a contract between a service provider and a customer that outlines the service to be provided, the level of performance to be expected, how performance will be measured and approved and what happens if performance levels are not met. SLAs are important for ensuring that both parties have a clear and shared understanding of the service expectations and responsibilities.

A composite SLA is the resulting SLA when combining multiple services that support an application, with differing levels of availability. The composite SLA can be higher or lower than the individual SLAs, depending on the application architecture and the logical operations (AND or OR) used to calculate it.

For example, consider an Azure App Service web app that writes to Azure SQL Database. Hypothetically, these SLAs might be:

App Service web apps = 99.95%
SQL Database = 99.99%

If either service fails, the whole application fails. The probability of each service failing is independent, so the composite SLA for this application is 99.95% × 99.99% = 99.94%. That's lower than the individual SLAs, which isn't surprising because an application that relies on multiple services has more potential failure points.

You can improve the composite SLA by creating independent fallback paths. For example, if SQL Database is unavailable, put transactions into a queue to be processed later. With this design, the application is still available even if it can't connect to the database. However, it fails if the database and the queue both fail at the same time. The expected percentage of time for a simultaneous failure is 0.0001 × 0.001, so the composite SLA for this combined path is:

Database or queue = 1.0 − (0.0001 × 0.001) = 99.99999%

The total composite SLA is:

Web app and (database or queue) = 99.95% × 99.99999% = ~99.95%

There are tradeoffs to this approach. The application logic is more complex, you are paying for the queue, and you need to consider data consistency issues.

I hope this helps you understand what SLAs and composite SLAs are and how to calculate them.

High Availability of two Azure Functions with Azure Traffic Manager

Wriju's Blog — Fri, 19 Jan 2024 16:29:12 +0000

How much Azure Open AI would cost?

Wriju's Blog — Fri, 08 Dec 2023 16:45:55 +0000

Azure OpenAI Service is a cloud-based platform that allows you to leverage large-scale, generative AI models using Azure infrastructure. The pricing for Azure OpenAI Service is based on a pay-as-you-go consumption model, which means you only pay for what you use. The price per unit depends on the type and size of the model you choose, as well as the number of tokens used in the input and output.

There are different types of models available in Azure OpenAI Service, such as language models, image models, embedding models, and speech models. Each model has a different price per unit, which you can find in the Azure pricing calculator. Some models also have additional costs for training, hosting, or inference.

For example, if you want to use the GPT-4 8K model, which is a large-scale language model, you would pay $0.03 per 1,000 tokens for the context prompt and $0.06 per 1,000 tokens for the completion. A token is roughly four characters for typical English text. You would also pay $102 per compute hour for training and $7 per hour for hosting the model.

If you want to use the Standard Dall-E model, which is an image model that can generate images from text, you would pay $2 per 100 images. There are no additional costs for training or hosting this model.

If you want to use the Standard Ada model, which is an embedding model that can create vector representations of text, you would pay $0.0001 per 1,000 tokens. There are no additional costs for training or hosting this model.

If you want to use the Whisper model, which is a speech model that can generate natural-sounding speech from text, you would pay $0.36 per hour. There are no additional costs for training or hosting this model.

To estimate your expected monthly costs for using Azure OpenAI Service, you can use the Azure pricing calculator and enter your expected usage details. You can also use the Azure Cost Management features to set budgets and monitor your actual costs. You can review forecasted costs and identify spending trends to optimize your spending.

Introducing Azure Co-pilot

Wriju's Blog — Wed, 06 Dec 2023 16:11:21 +0000

Azure Co-pilot is a preview service that uses AI to help you manage your Azure resources and applications. You can interact with Azure Co-pilot using natural language and ask questions, perform tasks, or get recommendations. Azure Co-pilot can also generate code, queries, and commands for you based on your goals and preferences. Azure Co-pilot is powered by large-scale, generative AI models that can understand and produce natural language and code. Azure Co-pilot is designed to be your AI companion and assistant for Azure operations and management.

A Guide to Programming Languages for Students

Wriju's Blog — Thu, 30 Nov 2023 16:36:56 +0000

Programming languages are the tools we use to instruct computers. Each language has its own strengths, weaknesses, and ideal use cases. Here's a guide to help you choose the right one for your needs.

Python

Python is a high-level, interpreted language known for its readability. It's great for beginners due to its simple syntax. Python is widely used in data analysis, machine learning, AI, web development, and automation.

print("Hello, World!")

Java

Java is a versatile, object-oriented language that's used in web development, mobile app development (especially Android), and large systems development. It's known for its "write once, run anywhere" philosophy.

public class Main {
    public static void main(String[] args) {
        System.out.println("Hello, World!");
    }
}

JavaScript

JavaScript is primarily used for enhancing web pages to provide for a more user-friendly experience. It's essential for front-end web development, and is increasingly being used on the server-side via Node.js.

console.log("Hello, World!");

C++

C++ is a high-performance language commonly used in system/software development, game development, and embedded systems. It's not as beginner-friendly as some languages, but it's powerful and widely used.

#include <iostream>
int main() {
    std::cout << "Hello, World!";
    return 0;
}

Swift

Swift is a language developed by Apple for iOS and macOS development. It's designed to be easy to use, and it's the fastest way to build apps for Apple platforms.

import Swift
print("Hello, World!")

Conclusion

The best programming language for you depends on what you want to do. Python or JavaScript are great for beginners, Java is widely used in large systems, C++ is for high-performance applications, and Swift is for Apple development. Happy coding!.

Different Types of Artificial Intelligence (AI)

Wriju's Blog — Tue, 28 Nov 2023 08:23:04 +0000

Artificial Intelligence (AI) is a broad field with several types and functionalities. It can be categorized based on capabilities and functionalities. Let us delve into these categories:

AI Based on Capabilities

1. Narrow AI

Narrow AI, also known as Weak AI, is designed to perform a narrow task such as voice recognition. These systems operate under a limited set of constraints and are focused on carrying out specific tasks. This is the reality today. What we see in the year 2023 and before does fall under this category.

The remaining two below are more of a theory and not yet consumable as a product.

2. Artificial General Intelligence (AGI)

AGI, often called Strong AI, is a type of AI capable of understanding, learning, and applying its intelligence to any intellectual task a human can do. It can understand, interpret, and respond to its environment as a human would. Just to clarify, AGI is not Generative AI.

3. Super AI

Super AI is an AI that surpasses human intelligence and ability. It can perform tasks currently not feasible for humans. This form of AI is not yet present but is the subject of interesting future discussions.

AI Functionalities

Under Narrow AI

1. Reactive Machine AI

Reactive Machine AI is the most basic type of AI system which cannot form memories or use past experiences to inform current decisions. It reacts to a set of inputs and performs a function.

2. Limited Memory AI

Limited Memory AI can make informed and improved decisions by using past experiences stored in its memory. It is capable of learning from historical data to make predictions.

Under AGI

1. Theory of Mind AI

Theory of Mind AI is a more advanced type of AI that can understand thoughts and emotions affecting human behavior. It understands that others have beliefs, desires, intentions, and perspectives that are different from their own.

Under Super AI

1. Self-aware AI

Self-aware AI is an AI that has evolved to be so akin to the human brain that it has developed self-awareness. Possessing consciousness, this type of AI does not currently exist.

AI is a fascinating and complex field, with many types and functionalities. As technology advances, we may see the emergence of new types of AI, expanding the possibilities of what AI can do.

What is the fundamental difference between Algorithm, Programming and Machine Learning Model?

Wriju's Blog — Mon, 27 Nov 2023 15:52:46 +0000

Algorithm: An algorithm is a set of well-defined instructions or procedures that are run on data to perform specific tasks. It can be described using math and pseudocode, and can be implemented with any one of a range of modern programming languages. Examples of algorithms include sorting algorithms like bubble sort, and search algorithms like best-first.

Programming: Traditional programming is a manual process where a person (programmer) creates the program. This involves formulating or coding rules based on the programmer’s knowledge and understanding of the problem. A computer program is a compiled form of human redable code which then executes based on the OS and build of the machine known has machine language. Programming is a step by step instruction given to a computer. Outcome of a program is deterministic.

Machine Learning Model: A machine learning model is an output of a machine learning algorithm, and it comprises of model data and a prediction algorithm. Machine learning algorithms perform automatic programming, where the models represent the programs. These models learn from data, or are “fit” on a dataset. Examples of machine learning algorithms include Linear Regression, Logistic Regression, Decision Tree, Artificial Neural Network, k-Nearest Neighbors, and k-Means.

In summary, an algorithm is a set of instructions used to perform a task, programming is the process of creating these instructions, and a machine learning model is the result of running a machine learning algorithm on data.