Tech Tips, Tricks & Trivia

This hour-long video covers not just the maths behind LLMs but also the history. Complex topics are well explained.  * Training a large language model is like constructing a skyscraper, where every   brick is placed by an army of specialists working around the clock. But here, the bricks are data, the mortar is mathematics, and the blueprint is a complex interplay of algorithms.  *   LLMs don't truly understand. They excel not through cognition, but through colossal computation. *  Mathematics is the invisible backbone of these 'thinking machines'. *  N-grams - A statistical method used by early language models that broke down text into small sequences of words to predict the next word based on common combinations in a dataset. *  Transformers - A groundbreaking architecture introduced in 2017 that revolutionized the field by enabling machines to grasp context at an unprecedented scale , allowing models to pay attention to all parts of a se...

In Episode #33 - Data Visualization Community of  Data Visualization, Data Storytelling, and Information Design - Lesson and Listen Series , Bill Shander highlights the importance of community in learning data visualization, acknowledging that while they didn't have formal training, they learned from the broader data visualization community. - Bill recommends various resources for data visualization, including: Books: There are many and Bill has shared a list of books that he has found useful and inspiring : Naked Statistics by Charles Wheelan, 2014, 304 pages Factfulness by Hans Rosling, Anna Rosling, and Ola Rosling, 2018, 352 pages Now You See It by Stephen Few, 2021, 301 pages The Functional Art by Alberto Cairo, 2012, 384 pages Storytelling with Data by Cole Nussbaumer Knaflic, 2015, 288 pages Information Visualization: Perception for Design by Colin Ware, 2020, 560 pages Good Charts: The HBR Guide to Making Smarter, More Persuasive Data Vis...

Harsh Dattani's ~60-minute walkthrough of Google AI Studio during the workshop " Build AI Apps using Google AI Studio & Gemini AI " provides all the essentials you need to get started. Topics covered: Google AI Studio -browser-based IDE for prototyping with generative models Gemini AI - lets you access the latest generative models from Google API key - prerequisite for building apps with Gemini AI Prompt engineering Prompt Gallery  - examples of what's possible with Gemini models Fine tuning  - provides a mechanism for curating output when you have a small dataset of input/output examples System Instructions - for developers to manage prompts  idx.dev - cloud-based IDE for building Web, Android apps Focus of the workshop is on the inner loop of software development Google AI Studio provides several interfaces for prompts that are designed for different use cases Google AI Studio is a good starting point for beginners and small-scale projects, while Vertex...

The 22-page free ebook " A Beginner's Guide to Data & Analytics " from Harvard Business School Online offers essential definitions and practical advice for those looking to learn more about the field. * Data science is the process of building, cleaning, and structuring datasets to analyze and extract meaning.  * Data analytics refers to the process and practice of analyzing data to answer questions, extract insights, and identify trends * Types of Analytics: Descriptive analytics looks at data to examine, understand, and describe something that’s already happened. Diagnostic analytics goes deeper than descriptive analytics by seeking to understand the “why” behind what happened. Predictive analytics relies on historical data, past trends, and assumptions to answer questions about what will happen in the future. Prescriptive analytics identifies specific actions an individual or organization should take to reach future targets or goals * Data literacy is the ability to...

Key takeaways from Episode #24 - Visualizing Health and Healthcare Data  of Bill Shander's Data Visualization, Data Storytelling, and Information Design - Lesson and Listen Series : Data visualization principles apply across industries, but domain expertise is crucial for effective communication. Healthcare data requires domain knowledge to ensure accuracy and helpfulness. Domain expertise is essential in healthcare data visualization due to its complexity and nuances. There are no homogeneous audiences, and healthcare data often involves multiple groups with competing points of view. Healthcare data often has life-and-death consequences, as seen during the COVID-19 pandemic. Proper scaling (linear vs. logarithmic) and clear communication of concepts (e.g., efficacy vs. effectiveness) are critical in healthcare data visualization. In the Listen section of the episode, Kathy Rowell, author of "Visualizing Health and Healthcare Data," shares insights on how to visualize hea...

Key takeaways from episode 23  Making Decisions with Data Visualization  of Bill Shander's Data Visualization, Data Storytelling, and Information Design - Lesson and Listen Series : Data visualization is crucial for enabling decision-making, as raw data can be boring and difficult to interpret. Understand the decision that needs to be made. Know what questions need answering Understand the data and its meaning in context . Know what your data is saying. The clearer you define questions, the more likely you'll gather and share data that helps your audience make the best decisions. Knowing the data and its relevance to the decision-making process is crucial, going beyond simply locating and calculating KPIs. Visualizations help to identify patterns and trends that may not be apparent in raw data, aiding in better decision-making. Creating visualizations is a consultative process The Listen part of the episode has an interview with Steve Wexler, an expert in data visualizati...

In the TED talk, AI Is Turning into Something Totally New , Microsoft AI CEO Mustafa Suleyman proposes that a more apt metaphor for AI is a new digital species .  I predict that we'll come to see them as digital companions, new partners in the journeys of all our lives. ...waves of history are clearly speeding up, as each one is amplified and accelerated by the last. And if you look back, it's clear that we are in the fastest and most consequential wave ever. The journeys of humanity and technology are now deeply intertwined.  ...today, the most advanced AIs consume more than eight trillion words in a single month of training. And all of this is set to continue.  ...just as the internet gave us the browser and the smartphone gave us apps, the cloud-based supercomputer is ushering in a new era of ubiquitous AIs. Everything will soon be represented by a conversational interface. Or, to put it another way, a personal AI. And these AIs will be infinitely knowledgeable, a...

Vertex AI is Google's end-to-end AI-ML Platform . Vertex AI Platform unifies the entire ML workflow from training to deployment. Source: Vertex AI home page Vertex AI provides several options for model training and deployment : AutoML lets you train tabular, image, text, or video data without writing code or preparing data splits. Custom training gives you complete control over the training process, including using your preferred ML framework, writing your own training code, and choosing hyperparameter tuning options. Model Garden lets you discover, test, customize, and deploy Vertex AI and select open-source (OSS) models and assets. Generative AI gives you access to Google's large generative AI models for multiple modalities (text, code, images, speech). You can tune Google's LLMs to meet your needs, and then deploy them for use in your AI-powered applications. Vertex AI offers the following Google proprietary foundation Gen AI model APIs : Gemini API - Multimodal da...

Paraphrased from the official documentation - IBM's acquisition of Red Hat for $34B in 2019 was the largest in software history Red Hat open source technologies provide a unified, enterprise-grade, hybrid solution for running workloads consistently across on-site and cloud infrastructures. Cloud-native development is an application development, deployment, and operational discipline spanning people, process, and technology to speed up high quality application production.  Cloud native development combines the modular architecture like microservices, containers and API-driven programmable orchestration automated workflows to enable continuous application delivery.  It changes the entire life cycle of how requirements are collaboratively incepted, coded, tested, and deployed. RHEL and Red Hat OpenShift provide an abstraction layer that is certified for all the major clouds. Kubernetes is the foundational orchestration tool used in Red Hat OpenShift to manage the container deploy...

Paraphrased notes from the Microsoft documentation - Azure Blueprints allow you to implement governance as code . Azure Blueprints  is a service that lets you define a repeatable set of Azure resources to implement your organization standards, patterns, and requirements .  You'll use Azure Blueprints in a declarative way to build and deploy new environments with a set of built-in components. You should think about Azure Blueprints as a package that you'll use to deploy and deliver governance as code . You can integrate it with your continuous integration and continuous deployment pipelines. Blueprints are JSON files , and are composed of the following artifacts: Role assignments Policy assignments Azure Resource Manager templates Resource groups These artifacts are embedded into the package, which can then be composed, versioned, and assigned to a management group containing multiple subscriptions, or assigned directly to a single subscription. Once assigned, the package w...

Paraphrased notes from the official Microsoft documentation - Azure Management Groups  help you organize your resources and Azure subscriptions. Management groups give you enterprise-grade management at scale Management groups are containers that help you manage access, policy, and compliance across multiple subscriptions.  Create these containers to build an effective and efficient hierarchy that can be used with Azure Policy and Azure Role Based Access Controls.  Each management group and subscription can only support one parent. Management Groups contain one or more subscriptions.  Inside of Subscriptions are Resource Groups.  Resource Groups belong to exactly one Subscription .  A Subscription can have many resource groups, but a resource group may belong to only one subscription.  Resource groups are containers that hold related resources for an Azure solution. Azure management groups support Azure role-based access control (Azure RBAC) for a...

Azure resources communicate privately, directly, and securely with each other over virtual networks. VNet is scoped to a subscription . You can implement multiple virtual networks within each Azure subscription and Azure region. VNet is scoped to a single region/location ; however, multiple virtual networks from different regions can be connected together using Virtual Network Peering. from Learn Azure in a Month of Lunches Your subnets should not cover the entire address space of the VNet. Plan ahead and reserve some address space for the future. If you create the smallest available subnet of /29 (with eight IP addresses), Azure will retain five addresses It is recommended you have fewer large VNets rather than multiple small VNets. This will prevent management overhead. Each virtual network is isolated from other virtual networks within each Azure subscription. Vnet does NOT span Regions, it can span availability zones. You can connect one VNet to another VNet using either Virtua...

*  Tanzu is the Japanese word for containers * With  "the biggest rearchitecture of vSphere in a decade", VMware is making vSphere capable of managing both virtual machines and Kubernetes (K8s) clusters in version 7 of vSphere.  *  Tanzu is a portfolio of products centered on K8s . Following VMware's acquisition of Pivotal in December 2019, Pivotal's PKS (Pivotal Container Service) K8s distribution is now known as VMware Enterprise PKS and is now part of Tanzu.  * In 2019, VMWare also acquired Bitnami, a specialist in packaging open-source applications for easy deployment to targets including K8s. This technology is now behind the Tanzu Application Catalog, which allows users to deploy commonly used applications and runtimes (such as node.js, Wordpess, MySQL or TensorFlow.) * Tanzu portfolio has ability to work across multiple K8s distributions and public clouds * VMware Tanzu Kubernetes Grid Integrated Edition [ link to PDF datasheet ] builds on Kuber...

Azure Container Registry is a managed Docker registry service based on the open-source Docker Registry 2.0.  Azure Container Registry is a registry of Docker and Open Container Initiative (OCI) images as well as related content formats, such as Helm charts, OCI artifacts, and images built to the OCI image format specification. Container Registry is private and hosted in Azure.  Like Docker Hub, Container Registry is organized around repositories that contain one or more images.  Container Registry lets you automate tasks such as redeploying an app when an image is rebuilt. All images stored in a container registry are encrypted at rest. Container Registry is highly scalable, providing enhanced throughput for Docker pulls that can span many nodes concurrently. The Premium SKU of Container Registry includes 500 GiB of storage. In addition to storing and hosting images, you can also use Container Registry to build images.  Instead of building an image yourself and pus...

Condensed noes from the official documentation - Azure Batch is a platform service for running large-scale parallel and high-performance computing (HPC) applications efficiently in the cloud.  Azure Batch schedules compute-intensive work to run on a managed pool of virtual machines, and can automatically scale compute resources to meet the needs of your jobs. It is suitable for scenarios requiring  large-scale parallel processing & massive computational power like financial risk modeling, 3D image rendering, media transcoding, and genetic sequence analysis. At a high level, Azure Batch involves a number of components working together.  Azure Batch account, which acts as a container for all Batch resources. The Batch workflow begins by uploading these data and application files to an Azure storage account associated with the Azure Batch account.  You then create a Batch pool with as many Windows or Linux virtual compute nodes as needed. They can also be au...

* Azure Firewall  is a managed, fully stateful firewall-as-a-service with built-in high availability and unrestricted cloud scalability that protects your Azure Virtual Network resources. * You can deploy Azure Firewall on any virtual network, but customers typically deploy it on a central virtual network and peer other virtual networks to it in a hub-and-spoke model. * Azure Firewall supports inbound and outbound filtering. Inbound protection is for non-HTTP/S protocols. For example RDP, SSH, and FTP protocols. *  The firewall, subnet, VNet, and the public IP address all must be in the same resource group . *  Azure Firewall needs a dedicated subnet like the Azure Application Gateway . * Azure Firewall is integrated with Azure Monitor for viewing and analyzing firewall logs. * The Azure Firewall service which provides network- and application-level protection across different subscriptions and virtual networks complements network security group fun...

Azure Backup offers multiple components that you download and deploy on the appropriate computer, server, or in the cloud .  The following matrix from the official documentation provides a quick summary of the options & key terms ( for a better view of the table, click here ) - Feature Azure Backup (MARS) agent System Center DPM Azure Backup Server Azure IaaS VM Backup Benefits Back up files and folders on physical or virtual Windows OS (VMs can be on-premises or in Azure) No separate backup server required. Application-aware snapshots (VSS - Volume Snapshot Service ) Full flexibility for when to take backups Recovery granularity (all) Can use Recovery Services vault Linux support on Hyper-V and VMware VMs Back up and restore VMware VMs using DPM 2012 R2 Application-aware snapshots (VSS) Full flexibility for when to take backups Recovery granularity (all) Can use Recovery Services vault Linux support on Hyper-V and...

Key points to note about the cheapest Azure B-Series VMs from various online hyper-linked sources: B-series burstable VMs are suitable for workloads that do not need the full performance of the CPU continuously or workloads typically having burstable performance requirements. Example use cases include development and test servers, low-traffic web servers, small databases, micro services, servers for proof-of-concepts, build servers . B-Series VMs run on hosts with Intel Haswell 2.4 GHz E5-2673 v3 processors or better processors. The E5-2673 v3 is what is used on the D_v2- and F-Series hosts, so the B-Series is a throttled FS-Series machine .  B-Series uses a similar system to the AWS T2 Instance. A B-Series virtual machine has a per-core base performance level, which is a percentage of each core. If the virtual machine remains under this level, it will earn credits up to a certain level. These credits can be burned to allow the virtual machine to burst beyond this per-core bas...

Notes compiled from official Azure documentation - When you create an Azure virtual machine (VM), you must create a virtual network (VNet) or use an existing VNet When creating a virtual machine using the Azure portal, the portal creates a network interface with default settings for you.  A network interface enables an Azure Virtual Machine to communicate with internet, Azure, and on-premises resources. Before creating a network interface, you must have an existing virtual network in the same location and subscription you create a network interface in. You assign IP addresses to a VM using a network interface. Once a network interface is created, you cannot change the virtual network it is assigned to. You can change the subnet the network interface is assigned to after it's created. The virtual machine you add the network interface to must also exist in the same location and subscription as the network interface. The virtual machine you attach a network i...

Application Gateway operates as an application delivery controller (ADC) as a service and offers features like: * Autoscaling is useful for workloads with variable traffic. Application Gateway's  fixed capacity mode is useful for scenarios with consistent and predictable workloads. * Layer 7 routing capabilities including cookie-based session affinity, round-robin load distribution, URL path/content-based routing, ability to host multiple websites. * Web application firewall (WAF) * Zone redundancy * Secure Sockets Layer (SSL/TLS) termination (not a best practice) and end-to-end SSL * Custom error pages * HTTP to HTTPS Redirection Application Gateway is a layer 7 load balancer , which means it works with web traffic only (HTTP, HTTPS, WebSocket, HTTP/2) unlike a Load Balancer which load balances traffic at layer 4 (TCP/UDP).  Application gateway supports hosting up to 20 websites behind a single gateway that could all be protected against web attacks w...