Driving Efficiency with MLOps & Microsoft Azure
The advancements in machine learning has more and more enterprises turning towards the insights provided by it. Data scientists are busy creating and fine-tuning machine learning models for tasks ranging from recommending music to detecting fraud. However, as is always the case with new technology, machine learning comes with its own set of challenges: Concept Drift – Accuracy of model degrades over time due to disparity in training data vs production data Locality – Pre-trained models’ accuracy levels change with changing demography/geography/customer Data Quality – Changes in data quality affect accuracy levels Scalability – Data scientists, while good at creating models, don’t necessarily have the skills to operationalize models at enterprise scale Process & Collaboration – A lot of models are developed and remain confined to sandboxes or within silos in an organization with no clearly defined process for the model lifecycle Model Governance – Most data science projects have no model governance – who can create models, who can deploy them, what datasets were used for training? Most ML projects do not have this defined clearly A typical ML model lifecycle Here’s what a machine learning model lifecycle looks like: What is MLOps According to Wikipedia, “MLOps (‘Machine Learning’ + ‘Operations’) is a practice for collaboration and communication between data scientists and operations professionals to help manage the production ML lifecycle. MLOps looks to increase automation and improve the quality of production ML while also focusing on business and regulatory requirements. MLOps applies to the entire lifecycle – from integrating with model generation (software development lifecycle, continuous integration/continuous delivery), orchestration, and deployment, to health, diagnostics, governance, and business metrics.” How is MLOps different from DevOps So, is MLOps just another fancy name for DevOps? Since machine learning is also a software system, most of the DevOps practices apply to MLOps too. However, there are some important differences: Team skills: a machine learning team usually has Data Scientists or/and ML researchers who may be excellent with different modeling techniques and algorithms but lack the right software engineering skills for building enterprise-grade production systems. Continuous Integration (CI) is not only about testing and validating code and components, but also testing and validating data, data schemas, and models. Continuous Deployment (CD) goes beyond deploying a package or service. It requires deploying an ML training pipeline that should automatically deploy another service (model prediction service). Continuous Testing is unique to ML systems, which is concerned with automatically retraining and serving the models. Monitoring – ML uses non-intuitive mathematical functions. It requires constant monitoring to ensure its operating within regulation and that the models are making accurate predictions. Implementing MLOps with Azure Machine Learning Tavant’s Manufacturing Analytics Platform (TMAP) is an analytics and machine learning-based platform that provides important business insights to our customers in the manufacturing domain especially Warranty. It is based on Azure and we use Azure Machine Learning’s MLOps features for managing our models’ lifecycle. Here’s a list of features that Azure provides for MLOps: Workspace – An Azure Machine Learning workspace is the foundational resource that is used to experiment, train and deploy machine learning models Development Environment – Azure ML provides multiple pre-configured ML specific VMs and computes instances. These come with most of the Machine Learning and Deep Learning libraries pre-installed and configured. One can also choose to create a local development environment if required Data Set – This step involves connecting to different data sources like Azure Blob Storage, Azure Data Lake, etc. and create a Machine Learning Data set. This Dataset can be used to access the data and its metadata when we create a Run Experiment & Runs – An experiment is a logical grouping of all the trials or runs. For each ‘Run’, you can log the metrics, images, data or enable logging. All these will be attached to the corresponding ‘Run’ under the Experiment. Compute Target – Creating a compute target helps you run your machine learning training. This compute target can be local or remote Azure GPU/CPU VMs Model Training – Azure ML already comes with multiple Estimators for Sklearn, Pytorch, Tensorflow and Keras. These Estimators helps you organize the ML training. Azure ML also has a capability to create Custom Estimators of your choice. All training logs, versions, and details will be logged in under the ‘Run’ of your experiment. Model Registry – Once the ML training is complete with different ‘Runs’ and you get the ‘best model’, the next step is to register the model in the Azure Model Registry. Model Registry maintains the model versions, descriptions of the model, model metadata, etc. Model Profiler – Before you deploy the model for real-time inference, profiling the infrastructure requirements for the model is very important. Profiling will give you a better understanding of how much minimum memory and CPU’s required for the model to give low latency and high throughput. Model Deployment – A model can be deployed to Azure Container Instances or Azure Kubernetes. This step involves providing which model and what version to deploy, its configurations, and the deployment configurations. Data Collection – It is used to capture real-time inputs and predictions from a model. It is used to analyze the performance of a model. Data Drift – It helps you understand the change in features, data quality issues, natural data shift, change in the relationship between features, etc. Conclusion MLOps is a must for enterprises using machine learning at scale. It allows for managing the complete model lifecycle including model governance and should be made mandatory for all Machine Learning projects. Azure Machine Learning provides has a great feature set for implementing MLOps. It does lack some of the advanced features like model lineage but one can always use dedicated MLOps platforms like MLflow or DotScience on Azure to bring in any missing features.
Manipulating Objects in an Image Through Self-Supervised Scene De-Occlusion
A well-recognized paper https://xiaohangzhan.github.io/projects/deocclusion/ from CVPR2020, introduces a complete framework for reproducing and recreating objects in a scene. It is a fascinating read, so in this article, we are providing commentary on the key aspects of this paper. To keep this article short, here we mainly present the structure of the framework, and specifically, we do not cover how the involved convolutional networks Scene de-occlusion decomposes an image, extracting cluttered objects in it into entities of individual intact objects. Orders and positions of the extracted objects can be manipulated to recompose new scenes (Partial Completion Networks, or PCNets) work. We will cover that in a subsequent article. The framework presented in this paper trains two distinct convolutional networks, PCNets. Both convolutional networks have slightly altered UNet architectures, (https://arxiv.org/abs/1505.04597), whose output is at a pixel level. The PCNet-C (PCNet-Content) network uses a partial convolution for completing the unoccluded object. The images below are taken from the excellent video on the project’s page and available in the paper. The input for this framework is an image with appropriate bounding boxes around objects. From this input, we determine, in this sequence (numbering matches the image at the end of the post): the ordering of objects (1), the complete un-occluded shape of each object (4), and finally the complete un-occluded color and pattern of the object (6). With the un-occluded object, we can perform a variety of tasks such as rearranging the objects in a photo while maintaining the correct object ordering and performing image inpainting on the background. The image inpainting step is not addressed here as the paper’s novelty is in their self-supervised object de-occlusion method. Both PCNet networks are trained by creating random occluding shapes and overlaying them onto the image. By generating the occluding shapes, we can train the networks through self-supervision. The training of the PCNet-M (PCNet-Mask) network consists of placing a random shape either in-front of or behind the target object (as determined from the original bounding box). In both cases, the model is trained to predict the original target mask. The second case is meant for regularization and is necessary to prevent the model from always assuming an object is being occluded. PCNet-C is trained to complete the portion of the target object that is occluded from the random occluding shape. Note that, as in the image below, the target object is not only occluded by the random shape but also a car (the black car in the bottom right of the image). Any attempt to account for this occlusion in the training process would require knowledge of object ordering and thus require a supervised framework. Although we disregard this cars occlusion, other non-occluded car objects in the dataset will allow our model to learn the true shape of a car. Importantly, the authors found that this training procedure generalizes to cases where there are multiple occluders, perhaps not overlapping with the pastry in the final image below. The framework procedure is described below (numbers match picture numbering): Recover the ordering of objects in the image using the PCNet-M We test for the ordering between two objects by selecting each as the ‘target’ and running the PCNet-M to find the amodal/un-occluded mask. If an object’s modal/occluded mask matches that of their amodal/un-occluded mask, they are not occluded by the other object. Retrieve all of the objects occluding(blocking) a given target object Generate two images for input into PCNet-M: A black and white image centered on the target object with the target, background, and union of occluding objects distinguished as in the picture below. A RBG image centered on the target object with the union of occluding objects greyed out. Use the PCNet-M to predict the amodal/un-occluded mask of the target object Generate two images for input into PCNet-C: A black and white image centered on the target object distinguishing between the target occluded object and the rest of the image. An RBG image centered on the target object with the difference of the amodal mask vs the modal mask greyed out. Use the trained PCNet-C to predict the amodal/un-occluded object.
Predicting Quality Assurance Code for Warranty Claims using AI
The Warranty claim process for equipment manufacturers requires their dealers to provide certain details about the machine part e.g. part number, date of manufacture, etc. Some of the key details provided by the dealers are the descriptions of the problems seen by the dealer, the potential root cause of the problem, and the possible solution. These descriptions map to certain codes as prescribed by the manufacturer and the dealer must pick the right codes and submit the claim for warranty. The list of these codes is usually long, running into many hundreds of codes, and many times, dealers end up entering the wrong code or select a default code like ‘Others’ or ‘Miscellaneous’. Once the claim reaches the manufacturer, their subject matter experts (SMEs) must validate the codes as part of the Quality Assurance process and end up spending a lot of time in correcting the erroneous entries made by the dealers. This results in a longer claim cycle and cost for the manufacturer. Bringing AI into Quality Assurance One of Tavant’s customers has been facing this challenge and was exploring the options to solve this using Artificial Intelligence (AI). So, the idea was to use AI to predict the right code based on the description entered by the dealer in real-time so that the data entered is correct and the SMEs don’t have to spend a lot of time in validating and fixing the codes resulting in better data quality and significant savings for them. The Challenge At first glance, it looks like the problem can be solved as a classification problem using machine learning. However, there were certain challenges to handle such as: Unreliable historical data – as mentioned above, the available claims data in the warranty system has a large proportion of incorrect codes, hence cannot be relied upon as training data. For instance, we found that there was a lot of imbalance in the data with a lot of data under the ‘Miscellaneous’ category code Descriptions entered by the dealers are free-form text and can vary in the style of language used The customer QA team recently formulated a set of new codes making the current codes in the data obsolete The Solution Natural Language Processing (NLP) has progressed leaps and bounds with state of the art changing every few months. This has mainly been driven by the rise of Deep Learning and specifically the usage of Word Embeddings – starting with Word2Vec to the current state of the art models like BERT, ELMO, etc. In word embeddings, words having the same meaning have similar representation and can also maintain the context in which the word is used, thus being able to differentiate between things like Apple, the company, and Apple, the fruit. We also approached the problem using Word Embeddings. However, to handle the mentioned challenges and to build a production-grade solution, a lot more was required. Below are some of the salient features of our solution: Data augmentation – there was very little data (~ 2 descriptions/code) provided by the experts from the customer – recall they had formulated a set of brand-new codes. We used NLP based data augmentation techniques to generate realistic descriptions Semantic similarity – we used word embeddings to find semantically similar descriptions and associated codes Continual Learning – the solution presents the dealers with top-3 predictions, and the dealer can select the right one from them. This allows the model to learn and evolve with more data Low latency – the models predict the code with sub-second latency this ensuring good user experience High scalability – the models are containerized using Docker and orchestrated using Azure Kubernetes Services (AKS), ensuring high scalability with an increase in workload The whole exercise would have been meaningless without reasonably high accuracy. We beat the expectations and achieved accuracy levels of almost 90%. This will improve further as more data comes into the system. The Benefits The ROI of the solution is significant with respect to the improvements in business metrics, as mentioned below: Reduction in time for the dealers to select the QA codes Better data quality as dealers can no longer assign a “default” code for the descriptions entered Reduction in time spent by the manufacturer SMEs in correcting the codes entered by dealers – from weeks to minutes Reduced claim processing time
Bringing the Power of Artificial Intelligence to the Media Industry
In the coming years, Artificial Intelligence (AI) and Machine learning (ML) are going to change lives so much that today’s science-fiction will be tomorrow’s reality. Earlier, humans made machines to reduce physical efforts required to do jobs. AI and ML are now giving brains to these machines and making them smart, which further reduces efforts. AI and ML have a significant impact on every industry. Automated transportation, predictive policing, intelligent gaming, enhanced health care, and smart homes are a few examples. Like other domains, digital Media and Entertainment industry has evolved in leaps and bounds with Artificial Intelligence and Machine learning. With the power of Artificial Intelligence, the media industry is becoming more interactive, personalized, and engaged. The technology has enabled media providers to personalize entertainment to unimaginable levels using streaming services, including on-demand movies, music, live telecasts, etc. These technologies provide many algorithms to eliminate buffering and low-quality playback, getting you the best quality content using intelligence from your ISP providers. Also, with concepts of data science, ML algorithms intelligently analyze unlimited streaming data to understand consumer’s viewing habits to offer more useful recommendations. Along with this, the sequence-to-sequence learning technique of Machine learning can translate your content from one language to another, from one writing style to another, allowing your content to reach distinct audiences more efficiently. Similarly, NLP (Natural Language Processing) algorithms help writing top trending news stories to decrease the time to produce new content. Shelley, an AI tool developed by MIT, allows users to write horror and fictional stories through deep learning algorithms. Following this trend, one could state that the creators of the next great content may not be human at all. Traditional media planning models have limited information on behaviors and purchasing patterns of the targeted audience. AI and ML can power numerous kinds of models, such as: Machine Learning-based pay-per-click campaigns Machine Learning-based content campaigns Machine Learning-based hyper-targeted email campaigns Efficient predictive data modeling and planning These ML models help media planners and marketers increase customer reach, improve the relevance of the audience, and create a great user experience. With extensive interactions and existing research, AI will soon influence market strategies, including business models and sales processes. In the future, Sales and Media planners will be assisted by AI agents who will monitor telephonic conversations in real-time. For example, such AI agents might infer from the client’s tone that the client is unhappy with the approach of that media planner or salesperson, and AI agents might help them decide the best possible strategy. In this way, AI could augment the capabilities of team members. With the endless possibilities offered by AI, we certainly have exciting times ahead.
Maximize Production Capacity with Prescriptive Analytics
Management Science is an approach to decision making based on the scientific method that makes extensive use of quantitative analysis. In today’s world, many use the terms management science, operations research, optimization, prescriptive analytics, decision science interchangeably. One of the most significant management science applications developed by the operation research (OR) group came about as a result of the deregulation of the airline industry in late 1970. Consequently, a number of low-cost Airlines were able to move into the market by selling seats at a fraction of the price charged by established carriers such as American airlines. OR group suggested offering different fare classes (discount and full fare) and, in the process, created a new area of management science. The OR group was able to achieve this by using forecasting and optimization techniques to determine how many seats to sell at a discount and how many seats to hold for full fare. Let us understand what decision making is and what are the two approaches by which we can take decisions. Problem-solving can be defined as the process of identifying the difference between the actual and desired situation and then taking action to resolve the difference. In contrast, Decision making is the term generally associated with the first five steps of the problem-solving process. Thus, the first step of decision making is to identify and define the problems. Decision making ends with the choice of an alternative, which is the act of making the decisions. An example of decision making can be a student who needs to decide which job to choose based on job evaluation data that leads to decision-making problems. Making a choice from the available alternatives is difficult. If the student decides that salary is the only criteria, then his decision will be referred to as single criteria decision making if he selects multiple options along with salary, like location and potential advancement then his decision will be multi-criteria decision-making. And in the real world, it becomes even more complex to solve. This leads to two approaches, called the Qualitative approach and Quantitative approach. The Qualitative Approach is based primarily on your judgment and experience. It includes your intuitive “feel” for the problem and is more of an art than science. A simple example is a manager taking decision for his company based on his experience. The Quantitative Approach is followed when the problem is complex, then Quantitative Analysis of the problem can be an important consideration for your final decision. Using the quantitative approach, an analyst will concentrate on the quantitative facts or data associated with the problem and develop mathematical expressions that describe the objective, constraint, and relationships. Linear programming is used when the objective function and the constraints of the problem can be expressed as linear equations of decision variables. Such comes under prescriptive analytics, which helps in providing the optimal solution to a problem. Traditionally, Operations Research (OR) techniques are used for finding the optimal solution to a problem. Many machine learning algorithms use optimization techniques such as gradient descent while solving a problem. Based on the above understanding, let us try to solve a product mix problem that comes under linear programming. “Maximize production capacity of a Manufacture” Suppose two products need to be manufactured: Tables and Chairs. To manufacture a product requires two resources: budget ($) and labor (man-hours) The resources required to manufacture each product and the total available amounts of each resource are given in the table below: Such Linear/Integer programming problems can have a significant influence on the profitability of organizations. Modern-day issues can have several millions or billions of decision variables and are solved using sophisticated software tools such as IBM CPLEX and FICO Xpress, GAMS. Such problems can also be solved using R (library ROI, lP solver, optimx), Python (library pulp), and even excel using solver. While experimenting on various tools like GAMS, Excel Solver, R, and Python, we identified optimal solution is 3 tables and 6 chairs. When using the Branch and Bound algorithm, there could be a scenario where multiple optimal solutions exist. We then leverage a qualitative approach in decision making. Reference: An Introduction to Management Science by Anderson Sweeney Williams Martin Business Analytics by Prof U Dinesh Kumar
Retaining your Customers with Seamless Warranty Resolution
In this era, companies are trying to stray away from the term of one-time customers. Competition has become immense, and organizations are pulling every string possible to ensure they can retain and satisfy their customers. Would providing the best product or having the best sales team solve the problem? They are all the fundamental part of the product’s sale, but the customers must have a healthy relationship with the organization even after the first sale. Here is when the warranty comes into the picture. Warranty is a written contract issued to the customer by the company, promising to repair or replace a product, if necessary. But providing a warranty alone is not enough. The company should follow a seamless warranty management practice that gives resolution with the best outcome and in the best possible timeframe. How does a seamless warranty resolution satisfy the customer? Customers are satisfied when they see a quick turnaround time for their warranty claims. Every touchpoint with the customer will be satisfactory if the warranty system is in place to give all the required information. The system should help the dealer/company make quick decisions regarding the customers’ warranty claims. How can a company achieve a quick turnaround time? Turnaround time is the time taken to resolve an issue. It is based on various variables, such as: Online questionnaire Parts availability Human resources Streamlined approval flows Failure identification Fraudulent detection Consistent systems Fast processing Configurable system When a company achieves this, it would be able to process a customer’s warranty claim with the fastest and best resolution. The company can do all this with their allocated budget if they can streamline the process. Warranty systems give all essential data about the defects and can predict the possible future failures from the past data. The failure information captured in the system can provide valuable information to set up future field actions or campaigns. These field actions, when set up efficiently, can make a huge difference to customer satisfaction. A robust warranty system can address the following in providing a seamless customer experience: Providing timely update; the vehicle warranty status Ensuring quick claim processing time Providing notification to the customers on any scheduled maintenance or field action which may be due Having a telematics connection with the vehicle to record the machine activities Enabling customer feedback through an online portal Reducing wait time for claim approval Providing access to required information for the team to engage the customer’s issues Wrapping up When an organization has a seamless warranty management in place, it has an added advantage over the competitors. So, it is clear to say, a seamless warranty system not only helps manufacturers in streamlining their internal processes but also helps in increasing customer retention and satisfaction. Reach out to at [email protected] in case you wish to gain more insights.
The Rise of Programmatic Advertising
Publishing ads in a heterogeneous media platform is a complex mechanism. Programmatic solutions address this problem through automated guaranteed, unreserved fixed rate, invitation-only auctions, and open auction transactions. According to IAB, in today’s digital supply chain, automation will continue to refine buying and selling processes and shift attention to higher-value marketing and advertising functions. Programmatic Ad Trends: The share of programmatic advertising in digital ad trade is growing across the globe. The stake of programmatic ads is two-third of online display advertising space, according to Zenith programmatic trend projections 2019. The potential of the programmatic landscape may depend on the following digital streams. pDOOH (Programmatic Digital out-of-home) media pDOOH is a medium to display ads in open/publicly accessible environment like Roadside digital billboards Outdoor signage Apartments lift lobby Airports Public waiting rooms Shopping centers etc. This medium not only reduces management cost but also influences customers with varied content. In 2019, industry revenue from billboards grew by 0.6%, which is more than $8.6 billion in revenue. pDOOH space is bright – more education and standardization with less fragmentation sets the ball rolling to long-term success. Voice Activated Advertising Amazon Alexa, Google Assistant, and Apple home are now becoming an integral part of our lives. These medium guides not only potential customers in the right direction, but also serves effective tailored ads. According to the VoiceBot.AI survey, 25% of respondents said that the daily grocery orders were mostly placed through voice-assistants. By looking at all these opportunities to generate more revenue, streaming audio platforms such as Soundcloud, Spotify, and Pandora have launched their applications with ad impressions. Also, the number of people to pitch is growing fast, with an estimated 1.8 million smart speakers sold last year and 15.1 million expected yearly sales by 2020, according to researcher strategy analytics. Advertisers are observing voice-activated advertising as the gateway to behavior-based targeted consumer engagement. Connected TV (CTV) Over-the-Top (OTT) or CTV is a wireless or ethernet connected TV that can stream internet content like video or audio. Advertisers are now primarily investing in connected TV (new inventory becomes available, with YouTube, Hulu, and Roku) dominating the landscape. According to the eMarketer survey, CTV advertising will surpass $10 billion by 2021. Demand-side platforms (DSP) are building up their ability to sell ads to TV networks programmatically, and, with the increased availability of high-speed(5G) wireless networks, mobile users are increasingly watching digital videos on their phones and tablets. After COVID-19, CTV advertising may rise more steeply as focus on e-trade and digitalization is increasing. Header Bidding In the ad world, most ads are picked and served on a publisher, based on priority; for example, the preference is given to pre-book an ad and then to others like RTB. Header bidding is a new technique, and this is different from priority. On a web page, the header contains page metadata and script code. To display ads, script code executes and sends auction request to all demand partners. Therefore, header bidder auction takes place in the header of the page even before the page gets loaded on the publisher. This is different from RTB, where auction occurred for those inventories which remain unsold after pre-book; therefore, if we compare quality-wise, header bidder auctions for inventories, could be considered premium as these are accessed ahead of pre-book ads. The Road Ahead As artificial intelligence is growing in the programmatic landscape, advertisers have now started to contemplate the future for better results in advertising through different solutions and media. All the above channels aid in reaching the targeted audience and allow surprising customers with innovation.
3 Reasons You Must Adopt AI-Based Quality Engineering
Regulations are constantly changing in the mortgage industry. Lenders are under continuous pressure to meet fast approaching deadlines on UCD and HMDA. The Uniform Closing Dataset (UCD) is a standard industry dataset enabling information on the CFPB’s Closing Disclosure to be communicated electronically. The first deadline of 25 September 2017 mandates lenders to deliver borrower data and Closing Disclosure in the UCD file. UCD improves loan quality through increased data accuracy and consistency. This is of interest to the GSEs as it enhances the loan’s eligibility for sale in secondary markets. The year 2018 brings updates to the HMDA. The new HMDA rule requires over 48 data points to be collected, recorded and reported. This includes multiple new data points and a few modified from the previous rule. New fields include credit scores, CLTV ratio, DTI ratio, detailed demographic data etc. The CFPB asserts that the changes improve the quality and type of data reported by financial institutions leading to greater transparency. The updated regulations bring a new set of challenges to lenders. Investments in technology systems and processes can potentially increase the ever-rising loan origination cost. Data privacy and security is another concern. With the increased number of data fields, protecting sensitive borrower information is a priority. Additional data can also be used in fair lending claims thereby increasing litigations risks and costs. Since 2008, the mortgage industry has been taking giant strides in improving data reporting and compliance standards. TRID rule impacted the industry at almost every point along the transaction, and UCD/ HMDA will change the way data is collected, recorded, reported and delivered. Over the years, Tavant’s mortgage expertise has helped lenders implement regulatory changes with cutting edge technologies. In 2015, we helped multiple lenders achieve TRID compliance ahead of schedule. In 2017, we are doing the same with HMDA and UCD. It’s time to achieve Accelerated Compliance with Tavant Testing. The countdown is on! To learn more about our testing solution please visit: UCD/HMDA Compliance Testing by Tavant
Five Essential Features of a Good OTT Streaming Service
The recent spurt seen in streaming activity will be one of the many things that COVID-19 will leave behind as a legacy. However, this super fest of content consumption across dozens of streaming apps was in play last year itself as if all of this were prescience. From March 9 to March 16, the total streaming time grew to 156.1 billion minutes per day in the United States, compared to 127.6 billion minutes during the last week of February, as per Nielsen. In March, streaming accounted for 23% of consumer TV viewing time, up from 21% in February and 14% a year ago. All the major platforms have shown adaptability and scalability in handling the increased workload. It is a testament to the idea of cloud engineering and modern application concepts such as microservices. The core architecture of the streaming applications has evolved a long way over the last decade to be able to be so flawless now. This blog post tries to explore the concepts that are fundamental to building steady data streaming applications. There are many functional and performance requirements for which an OTT service must be ready: Sensitivity to the bandwidth disparities in different networks and the fluctuations in network bandwidth Performance latency and buffering issues Ever-increasing catalog of content to be made searchable and accessible Access to vast content on a site or app that is lightweight Prevent latency issue es for popular and most liked content Handle high load and ability to deliver to all requests Maintain consistency of the content usage across devices and time zones Security and data encryption as well as entitlement Ability to store and record billions of user actions and then process and use the insight to make the service better Intuitive user experience across different screen and device sizes These are but some of the requirements that a good streaming service would require to fulfill. While there are too many things from an architecture perspective that go into the building of a streaming app, we will try to cover a few foundation elements. Edge Computing and CDNs: Content is delivered to devices through CDNs, which act as the cache for the content. Edge computing uses the power of the cloud and takes computing power close to the end device. This is the first layer of logic that interacts with devices, and it links the request to the appropriate API within the services architecture. It also provides the abstraction layer to the mid-tier services. Load Balancing: This helps the streaming service manage peaks in the load by implementing throttling mechanisms that reject the extra incoming requests and diverting them to other servers when the traffic crosses a threshold. Microservices: The entire backend application is split into hundreds of independent services. These services implement specific business logic, hence encapsulating one service from another. This enables the entire service to function without the risk of one flaw bringing the whole application down to its knees. These services provide functions such as authentication, licensing, playback, artwork, etc. Encoding and Content Delivery: Every media file is broken into chunks and transcoded into different bit rates. This is done to provide the best possible quality to different devices at varying bandwidths. Every media content demands its quality standards. A fast-moving video will require transcoding at a higher quality than a video, which is relatively slow-moving. Adaptive streaming is used to push the most appropriate bit rate segment of the video. Local caching is used to serve subscribers. Push fill methodology is used to load the content in a regional CDN based on the popularity of the content in that region. Data pipelines: Data is created at an enormous scale since billions of events take place in a day and millions in a second. A data pipeline must be designed to enable the cloud storage of the video viewings and UI activities. This data is used for big data analytics. This, along with device error logs and diagnostic events, is also used for monitoring and debugging operations. Final Thoughts Like Rome was not built in a day, it takes time to build excellent streaming services. Most of these components of a service evolve over time. The three goals of any streaming service to achieve architectural excellence are scalability, availability, and immutability. Scalability enables the streaming service to scale for massive audiences and handle spikes efficiently. Availability allows the service to be available for requests while maintaining the site’s performance. Immutability provides robustness to the architecture so that any change to the infrastructure on the fly does not upset the system and disrupt the service. Companies can use third party tools or build the components on their own, depending on the scale of the service they are trying to develop. Custom build applications provide more control to the companies, and they can build services based on their requirements. Several open-source components can be used, which can be then built upon to create exciting, viable products that can scale quickly.
Aftermarket 4.0 During and Post COVID-19 (Using IoT, AI, Chatbot and DIY)
In today’s volatile, uncertain, complex, and ambiguous world, we are facing unexpected impediments. And manufacturers are one of the most negatively impacted industries. Due to plant shutdown, Auto manufacturers alone in India incurred a loss of close to three billion dollars, and it continues.[i] David Nabarro, WHO special envoy for COVID-19, stated, “I believe that the world will have to learn to live with Corona in our midst, and that means being on a constant defense everywhere as we do with other infectious diseases.” [ii] Businesses that want to survive and win are looking out for avenues to create a system of disaster-response infrastructure. They understand that most customers are not loyal to a brand; they are loyal to their needs of that brand. “Needs” such as quality, prompt response, timely services, value for money, or priority privilege, and so forth. Organizations are finding ways to reach their customers and reassure their peace of mind. To offer initial diagnostic service call, Consumer Priority Service Corporation, an international provider of repair and extended warranty services for the consumer electronics and appliance industries, has partnered with Zoom.[iii] Warrantywise, a British auto extended Warranty selling company, is now allowing their dealers to sell extended warranty policies from home.[iv] Almost all auto companies are offering an extension to their policies that are about to expire soon. Is that enough to keep up the brand performance? How do the leaders ensure the safety of their employees and customers and still secure the supply chain efficiency? How can they manage near-term revenue and cash expenses while satisfying endless customer needs at the same time? Can sales be achieved as a by-product of responsive, reliable, and reassuring warranty services offered to the customers who are dealing with much uncertainty? Dealing with such a situation, the leaders must equip their companies with a pandemic-response capability by building a robust digital infrastructure with the help of Aftermarket 4.0. So, what is Aftermarket 4.0? It is an Aftermarket services version of Industry 4.0. Aftermarket 4.0 is an infrastructure and practice that includes cyber-physical systems, IIoT, cloud computing, big data, and predictive analytics to instill a culture and trend of automation and data exchange in the operations. The benefits are enormous, such as: 1. Data Access across the supply chain 2. Improved safety, output, and satisfaction for employees, workers, and customers 3. Enhanced customization based on customer and operational needs and different services offered 4. Improved productivity and business output To enable these changes as a leader, you must follow the five points described below: 1. An Intelligent Warranty Management System: First thing first, all manufacturers that are in the business of warranty must go for a seamless Warranty Management system to automate complex human-led processes and decisions. Manufacturers must get rid of all their age-old on-premise solutions. They must aim for an intelligent cloud-based system that seamlessly binds different organizational aspects, both internally and externally. 2. Building AI Base: Leaders must consider the capability of the warranty management system to working seamlessly with near real-time data received over IoT and uses artificial intelligence to automate more decisions that were otherwise needed collaborative human interference. 3. Enabling Chatbot: By consulting and evaluating the historical customer queries, the AI system can identify and present various patterns of problems vs. solutions. Chatbots, backed by extensive data, can especially be very useful in responding to customers’ queries practically instantaneously. This investment towards the Chatbot can save the enormous human hours spent otherwise. And, at the same time, reassure customers with prompt and significant responses. 4. Prescriptive Maintenance: It is no longer enough to predict the failures alone. The ‘Smart Manufacturing’ demands ‘Prescription’ along with the ‘Prediction’. Prescriptive maintenance not only tells you that a problem is likely to occur but also shows you multiple response scenarios to choose from. By utilizing artificial intelligence and machine learning methods, perspective maintenance advises the technicians on what to do and how to perform a repair. 5. DIY Friendly products and solutions: The last step towards building the eco-system of robust customer satisfaction is to enable and offer more ‘Do-It-Yourself’ activities on maintenance for your customers. Encouraging your customers to use the mobile app version of your warranty management system to scan parts receipts/invoices to the appropriate places is a great way to replace the paper trail for the repair history, helping you to keep the warranty intact. Final Thoughts The clock is ticking. We are in unprecedented times. Nevertheless, there are many possibilities, but are any of them being practiced? And where do you get such a futuristic system? Reach out to us at [email protected] today. We got you covered! References: [i] https://www.deccanherald.com/business/business-news/coronavirus-impact-auto-industry-to-suffer-loss-of-rs-21000-crore-due-to-plant-closures-817266.html [ii] https://www.indiatoday.in/india/story/coronavirus-new-reality-we-will-have-to-learn-to-live-with-it-who-official-david-nabarro-1660190-2020-03-27 [iii] https://www.cpscentral.com/consumer-priority-service-partners-with-zoom/ [iv] https://cardealermagazine.co.uk/publish/warrantywise-can-help-dealers/188765