Categories
AI

Iterable optimizes AI to hyper-personalize marketing and predict future purchases

Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.


With people living on their phones and constantly deluged by messages, emails and notifications, so-called “batch-and-blast” emails, targeted web ads and word-of-mouth are no longer adequate when it comes to customer outreach.

“We’re online more than ever before, acutely aware and concerned with how our personal data is used and, importantly, we demand to be valued and recognized for what makes us unique,” said Andrew Boni, CEO of customer engagement platform company Iterable

Marketers must then achieve the “extraordinary task” of reaching as many customers — and in as many hyper-personalized ways — as possible. And, matching each consumer with “a personal marketer and targeted message at every touchpoint” is impossible without artificial intelligence (AI) and machine learning (ML), said Boni. 

To this end, Iterable today announced its new AI Optimization Suite. The tool is equipped with new predictive goals with explainable AI capabilities to help drive individualized brand-consumer communication, said Boni. 

The company will officially unveil its new platform to the public at its Activate Summit North America in September. 

“Understanding the inner workings [the ‘why’ and ‘how’] behind the system gives marketers the clarity and confidence to improve and refine the predictions they create, design goals that fit their unique business needs, and implement fresh ideas about how to approach customer-first campaigns,” said Bela Stepanova, senior vice president of product at Iterable. 

Iterable treats customers as its most valuable asset

Most brands have access to customer data and a goal in mind for connecting with their customers. But the magic formula is often missing: Building true customer knowledge — and acting on that. 

This has provided a significant opportunity and rapid growth in AI-powered customer engagement platforms. According to Markets and Markets, the global customer engagement solutions market will grow to $32.2 billion by 2027, representing a compound annual growth rate (CAGR) of nearly 11%. 

The global market for retail omnichannel commerce platforms will reach $16.9 billion by 2027, as forecasted by ReportLinker.com. This represents a CAGR of 16.4% from 2020 — growth largely propelled by the COVID-19 crisis, according to the firm. 

Companies competing for that market share include Adobe Marketo Engage, HubSpot Marketing Hub, Pega, Blueshift, Braze and MoEngage. 

And Iterable is rapidly increasing its presence: The 9-year-old series E “centaur” company that now serves more than 1,000 customers — including DoorDash, Calm, Jersey Mike’s Subs, Zillow and SeatGeek — has surpassed $100 million in annual recurring revenue (ARR), and has recently expanded into Latin America and Asia Pacific territories. 

Ingest, centralize, activate

Iterable’s AI Optimization Suite helps brands “ingest, centralize and activate” customer data and use real-time information to craft individualized campaigns, said Boni. 

Its predictive goals functionality allows brands to establish goals unique to their business needs and predict how their customer base might respond. They can then use those predictions to establish customer segments and tailor messaging to maximize conversion. 

The explainable AI feature within the predictive goals functionality allows marketers to pinpoint the data points that contributed to forecasts. This provides an “under the hood” look at AI and further insights that can inform future campaigns, said Boni.

Furthermore, marketers can understand behaviors that drive predictions, ranked in order of importance, according to Boni. They can also assess the quality and reliability of a goal using a predictive strength tool and receive insight into correlated variables that make outcomes more or less likely. 

Building a ‘What to Do’ journey

As an example: A travel company is looking to drive incremental purchases and increase rental car revenue. 

Predictive goals creates a list of users most likely to book a rental car within the next 30 days. Brands can then score that list and select and target the highest propensity users with cross-channel messages triggered after an airline ticket purchase, explained Boni. 

Using Explainable AI, the travel company can then go a step further by identifying the attributes that target those high-propensity customers — browsing a “Tours and Activities” section of their website after booking a flight, for instance. These gleaned insights can be used to create a “What to Do” journey for airfare purchasers. This might highlight tours and other activities in their travel destinations, wrapped together with rental car promotions, said Boni. 

The company plans to expand the platform to help marketers understand how goals perform over time, he said, as well as allowing them to more deeply analyze cohorts and optimize individual user experiences with journeys and experimentation.

“By setting goals and measuring outcomes, marketers can lean into strategies that are working and adjust any that aren’t,” said Boni. 

He marveled that, “we can make predictions about the future. This is where the world is heading. It is less about looking into the past and making a guess, and more about (establishing) future behavior.”

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.

Repost: Original Source and Author Link

Categories
AI

Microsoft’s Tutel optimizes AI model training

Hear from CIOs, CTOs, and other C-level and senior execs on data and AI strategies at the Future of Work Summit this January 12, 2022. Learn more


Let the OSS Enterprise newsletter guide your open source journey! Sign up here.

Microsoft this week announced Tutel, a library to support the development of mixture of experts (MoE) models — a particular type of large-scale AI model. Tutel, which is open source and has been integrated into fairseq, one of Facebook’s toolkits in PyTorch, is designed to enable developers across AI disciplines to “execute MoE more easily and efficiently,” a statement from Microsoft explained.

MoE are made up of small clusters of “neurons” that are only active under special, specific circumstances. Lower “layers” of the MoE model extract features and experts are called upon to evaluate those features. For example, MoEs can be used to create a translation system, with each expert cluster learning to handle a separate part of speech or special grammatical rule.

Compared with other model architectures, MoEs have distinct advantages. They can respond to circumstances with specialization, allowing the model to display a greater range of behaviors. The experts can receive a mix of data, and when the model is in operation, only a few experts are active — even a huge model needs only a small amount of processing power.

In fact, MoE is one of the few approaches demonstrated to scale to more than a trillion parameters, paving the way for models capable of powering computer vision, speech recognition, natural language processing, and machine translation systems, among others. In machine learning, parameters are the part of the model that’s learned from historical training data. Generally speaking, especially in the language domain, the correlation between the number of parameters and sophistication has held up well.

Tutel mainly focuses on the optimizations of MoE-specific computation. In particular, the library is optimized for Microsoft’s new Azure NDm A100 v4 series instances, which provide a sliding scale of Nvidia A100 GPUs. Tutel has a “concise” interface intended to make it easy to integrate into other MoE solutions, Microsoft says. Alternatively, developers can use the Tutel interface to incorporate standalone MoE layers into their own DNN models from scratch.

A line graph comparing the end-to-end performance of Meta’s MoE language model using Azure NDm A100 v4 nodes with and without Tutel. The x-axis is the number of A100 (80GB) GPUs, beginning at 8 and going up to 512, and the y-axis is the throughput (K tokens/s), beginning with 0 and going up to 1,000 in intervals of 100. Tutel always achieves higher throughput than fairseq.

Above: For a single MoE layer, Tutel achieves an 8.49 times speedup on an NDm A100 v4 node with 8 GPUs and a 2.75 times speedup on 64 NDm A100 v4 nodes with 512 A100 GPUs, Microsoft claims.

“Because of the lack of efficient implementations, MoE-based models rely on a naive combination of multiple off-the-shelf operators provided by deep learning frameworks such as PyTorch and TensorFlow to compose the MoE computation. Such a practice incurs significant performance overheads thanks to redundant computation,” Microsoft wrote in a blog post. (Operators provide a model with a known dataset that includes desired inputs and outputs). “Tutel designs and implements multiple highly optimized GPU kernels to provide operators for MoE-specific calculation.”

Tutel is available in open source on GitHub. Microsoft says that the Tutel development team will “be actively integrating” various emerging MoE algorithms from the community into future releases.

“MoE is a promising technology. It enables holistic training based on techniques from many areas, such as systematic routing and network balancing with massive nodes, and can even benefit from GPU-based acceleration. We demonstrate an efficient MoE implementation, Tutel, that resulted in significant gain over the fairseq framework. Tutel has been integrated [with our] DeepSpeed framework, as well, and we believe that Tutel and related integrations will benefit Azure services, especially for those who want to scale their large models efficiently,” Microsoft added.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
AI

Google’s Model Search automatically optimizes and identifies AI models

Google today announced the release of Model Search, an open source platform designed to help researchers develop machine learning models efficiently and automatically. Instead of focusing on a specific domain, Google says that Model Search is domain-agnostic, making it capable of finding a model architecture that fits a dataset and problem while minimizing coding time and compute resources.

The success of an AI model often depends on how well it can perform across various workloads. But designing a model that can generalize well can be extremely challenging. In recent years,”AutoML” algorithms have emerged to help researchers find the right model without the need for manual experimentation. However, more often than not, these algorithms are compute-heavy and need thousands of models to train.

Model Search, which is built on Google’s TensorFlow machine learning framework and can run either on a single machine or several, consists of multiple trainers, a search algorithm, a transfer learning algorithm, and a database to store evaluated models. Model Search runs training and evaluation experiments for AI models in an adaptive and asynchronous fashion, such that all trainers share the knowledge gained from their experiments while conducting each experiment independently. At the beginning of every cycle, the search algorithm looks up all the completed trials and decides what to try next, after which it “mutates” over one of the best architectures found up to that point and assigns the resulting model back to a trainer.

To further improve efficiency and accuracy, Model Search employs transfer learning during experiments. For example, it uses knowledge distillation and weight sharing, which bootstraps some of the variables in models from previously-trained models. This enables faster training and by extension opportunities to discover more and ostensibly better architectures.

After a Model Search run, users can compare the many models found during the search. In addition, they can create their own search space to customize the architectural elements in their models.

Google Model Search

Above: An example of an evolution of a model over many experiments. Each color represents a different type of architecture component.

Image Credit: Google

Google says that in an internal experiment, Model Search improved upon production models with minimal iterations, particularly in the areas of keyword spotting and language identification. It also managed to find an architecture suitable for image classification on the heavily-explored CIFAR-10 open source imaging dataset.

“We hope the Model Search code will provide researchers with a flexible, domain-agnostic framework for machine learning model discovery,” Google research engineer Hanna Mazzawi and research scientist Xavi Gonzalvo wrote in a blog post. “By building upon previous knowledge for a given domain, we believe that this framework is powerful enough to build models with the state-of-the-art performance on well studied problems when provided with a search space composed of standard building blocks.”

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
AI

Granulate raises $30 million for AI that optimizes server workloads

Granulate, a startup developing a platform that optimizes computing infrastructure, today announced it has raised $30 million, bringing its total raised to $45.6 million. The company claims the cash infusion will help it develop products that reduce the time engineers spend tuning enterprise system performance, freeing them up to pursue potentially more creative, impactful, and challenging projects.

Applying AI to datacenter operations isn’t a new idea — IDC predicts that 50% of IT assets in datacenters will run autonomously using embedded AI functionality by 2022. To this end, Concertio, which recently raised $4.2 million, provides AI-powered system optimization tools for boosting hardware and software performance. Facebook has said it employs AI internally to speed up searches for optimal server settings. And IBM offers a tool called Data Center Advisor with Watson that calculates datacenter health scores and anticipates possible issues or failures.

But Granulate’s solution uses software agents that can be installed on any Linux server in datacenters or cloud environments, including virtual machines. These AI-powered agents adapt to operating systems and kernels, prioritizing threads while taking into account each request’s processing stage and employing a network stack that enables parallelism. Granulate analyzes usage patterns and sizes to tailor allocations and memory for each app, autonomously crafting congestion control prioritizations between connections to optimize throughput for the current workload.

Granulate

Above: Above: Granulate’s analytics dashboard. Image Credit: Granulate

Image Credit: Granulate

Granulate’s suite, which works with existing monitoring tools like Prometheus, AppDynamics, New Relic, Datadog, Dynatrace, and Elastic, is installed in dozens of production environments and tens of thousands of servers, including those owned by Coralogix and AppsFlyer. The company claims it’s more performant than most, improving the throughput of machines by up to 5 times and leading to an up to 60% compute cost savings and a 40% reduction in latency.

Startapp, a mobile data platform with over 1.5 billion monthly active users and 800,000 partner apps, reports that Granulate achieved a 30% reduction in average latency and a 25% processor utilization reduction, netting a 33% compute cost reduction. Another customer — advertising technology company Bigabid — says it managed to reduce compute costs by 60% within 15 minutes of deploying Granulate.

In October, as part of a new partnership with Microsoft Azure, Granulate launched a real-time continuous optimization solution for Azure cloud and hybrid environments in the Microsoft Azure Marketplace. Earlier in the year, Granulate announced support for software-as-a-service contracts in AWS Marketplace, enabling Amazon Web Services customers to prepay for Granulate based on expected usage tiers through contracts up to two years long.

Granulate claims that in the past 10 months, its 21 customers have saved over 3 billion hours of core usage, with the number of CPU cores under management rising by over 10 times to over 300,000 cores. It expects 2021 revenue to hit $5 million, up from between $1 million and $2 million in 2020.

“The pandemic created a large market opportunity for us, as companies of all shapes and sizes began implementing solutions that could reduce costs. Our unique optimization solution enables companies to achieve significant computing cost reduction and performance improvement on any infrastructure or environment with minimal effort or code changes,” CEO Asaf Ezra, who cofounded Granulate in 2018 with Tal Saig, told VentureBeat via email. “We look forward to continuing to help customers of all sizes and industries as we develop new solutions to drive further improvements in computing efficiency.”

Granulate’s series B round was led by Red Dot Capital Partners with the participation of Insight Partners, TLV Partners, and Hetz Ventures. Dawn Capital also joined as a new investor.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform
  • networking features, and more

Become a member

Repost: Original Source and Author Link