Categories
AI

AI model optimization startup Deci raises $21M

Tel Aviv, Israel-based Deci, a company developing a platform to optimize machine learning models, today announced that it raised $21 million in a series A round led by Insight Partners with participation from Square Peg, Emerge, Jibe Ventures, Samsung Next, Vintage Investment Partners, and Fort Ross Ventures. The investment, which comes a year after Deci’s $9.1 million seed round, brings the company’s total capital raised to $30.1 million and will be used to support growth by expanding sales, marketing, and service operations, according to CEO Yonatan Geifman.

Advancements in AI have led to innovations with the potential to transform enterprises across industries. But long development cycles and high compute costs remain roadblocks in the path to productization. According to a recent McKinsey survey, only 44% of respondents reported cost savings from AI adoption in business units where it’s deployed. Gartner predicts that — if the current trend holds — 80% of AI projects will remain “alchemy,” run by “[data science] wizards” whose talents “will not scale in the organization.”

Deci was cofounded in 2019 by Geifman, Ran El-Yaniv, and entrepreneur Jonathan Elial. Geifman and El-Yaniv met at Technion’s computer science department, where Geifman was a PhD candidate and El-Yaniv a professor. By leveraging data science techniques, the team developed products to accelerate AI on hardware by redesigning models to maximize throughput while minimizing latency.

“I founded Deci in 2019 with Professor Ran El-Yaniv and Jonathan Elial to address the challenges stated above. With our talented team of deep learning researchers and engineers, we developed an innovative solution — using AI itself to craft the next generation of AI. By utilizing an algorithmic-first approach, we focus on improving the efficacy of AI algorithms, thus delivering models that outperform the advantages of any other hardware or software optimization technology,” Geifman told VentureBeat via email.

Deci achieves runtime acceleration on cloud, edge, and mobile through data preprocessing and loading, automatically selecting model architectures and hyperparameters (i.e., the variables that influence a model’s predictions). The platform also handles steps like deployment, serving, and monitoring, continuously tracking models, and offering recommendations where customers can migrate to more cost-effective services.

“Deci’s platform offers a substantial performance boost to existing deep learning models while preserving their accuracy,” the company writes on its website. “It designs deep models to more effectively use the hardware platform they run on, be it CPU, GPU, FPGA, or special-purpose ASIC accelerators. The … accelerator is a data-dependent algorithmic solution that works in synergy with other known compression techniques, such as pruning and quantization. In fact, the accelerator acts as a multiplier for complementary acceleration solutions, such as AI compilers and specialized hardware.”

AutoNAC

Machine learning deployments have historically been constrained by the size and speed of algorithms, as well as the need for costly hardware. In fact, a report from MIT found that machine learning might be approaching computational limits. A separate Synced study estimated that the University of Washington’s Grover fake news detection model cost $25,000 to train in about two weeks, and Google spent an estimated $6,912 training BERT.

Deci

Above: Deci’s backend dashboard.

Image Credit: Deci

Deci’s solution is an engine — Automated Neural Architecture Construction, or AutoNAC — that redesigns models to create new models with several computation routes, optimized for an inference device and dataset. Each route is specialized with a prediction task, and Deci’s router component ensures that each data input is directed via the proper route.

“[O]ur AutoNAC technology, the first commercially viable Neural Architecture Search (NAS), recently discovered DeciNets, a family of industry-leading computer vision models that have set a new efficient frontier utilizing only a fraction of the compute power used by the Google-scale NAS technologies, the latter having been used to uncover well-known and powerful neural architectures like EfficientNet,” Geifman said. “Such models empower developers with what’s required to transform their ideas into revolutionary products.”

The thirty-employee company, Deci, recently announced a strategic collaboration with Intel to optimize AI inference on the chipmaker’s CPUs. In addition to Intel, the startup says that “many” companies in autonomous vehicle, manufacturing, communication, video and image editing, and health care have adopted the Deci platform.

“Deci was founded to help enterprises maximize the potential of their AI-based solutions. Enterprises that are leveraging AI face an upward struggle, as research demonstrates that only 53% of AI projects make it from prototype to production,” Geifman said. “This issue can largely be attributed to difficulties navigating the cumbersome deep learning lifecycle given that new features and use cases are stymied by limited hardware availability, slow and ineffective models, wasted time during development cycles, and financial barriers. Simply put, AI developers need better tools that examine and address the algorithms themselves; otherwise, they will keep getting stuck.”

Deci has competition in OctoML, a startup that similarly purports to automate machine learning optimization with proprietary tools and processes. Other competitors include DeepCube, Neural Magic, and DarwinAI, which uses what it calls “generative synthesis” to ingest models and spit out highly optimized versions.

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

Cloud optimization startup Cast AI raises $10M

Cast AI, a cloud optimization and management platform, today announced that it raised $10 million in series A funding led by Cota Capital, with Samsung Next and additional investors participating. CEO Yuri Frayman says that the funding will be put toward expanding its sales team and further developing its AI-powered platform.

Cloud adoption exploded during the pandemic, with Forrester forecasting that the public cloud infrastructure market will increase 28% year-over-year to $113.1 billion in 2021. But as enterprises increasingly embrace cloud technologies, they’re devoting larger-than-anticipated portions of their IT budgets to cloud spend. According to one survey, 55% of executives were surprised by cloud costs or experienced sudden cost spikes — many of which weren’t discovered until days, weeks, or even months later.

Based in Miami, Florida, and founded in 2019, Cast leverages AI trained on millions of data points to balance cloud performance and cost. Supporting Amazon Web Services, Google Cloud Platform, and Microsoft Azure, the platform connects to existing clouds and generates a report to identify cost-saving opportunities.

“Cast was created because, during the founders’ previous venture, Zenedge, which was later acquired by Oracle, they experienced enormously high cloud bills and found it hard to understand how to better manage this expense,” Frayman, who cofounded the company with Leon Kuperman and Laurent Gil, told VentureBeat. “They quickly realized that many other business leaders were having similar struggles and that managing cloud is a complex and time-consuming task requiring a lot of expertise. They wanted to lift the burden for other companies, so their engineers could focus on building what they love, and not worrying about infrastructure managing ups and downs.”

Cloud spend analysis

Even before the pandemic, many enterprises were migrating their software operations to the cloud. A 2018 IDG survey found that 73% of organizations had at least one app or a portion of their computing infrastructure already in the cloud. But the challenges both then and today are myriad, with companies responding to a Lemongrass survey by citing security, compliance, costs, and a lack of in-house skills as the top barriers to adoption.

To address the cost concerns, Cast employs a number of heuristics and trained machine learning models to help predict the future state of customers’ apps for the purposes of scaling. For example, the platform has models that attempt to be proactive with the scaling requirements for each cloud cluster under management. If the models — which are customer-specific — predict incorrectly, Cast uses the errors as additional data points to improve the model.

“We have other models that use global datasets for market characteristic predictions,” Frayman explained. “For example, we train a global model to predict instance preemptions by machine type, region, availability zone, and seasonality. This model is shared autonomously across all customers, and all the data is used to retrain the model continuously. One of the variables of our global model started by benchmarking more than 500 different instance types in four cloud providers in all regions across the world. This helps the global model to understand and contrast instances and how much compute quantity you get per virtual machine type.”

Cast competes with Kubecost, Spotinst, and Granulate in the cloud optimization solutions market, among others. But the company claims to have more than two dozen customers across the adtech, ecommerce, fintech, and data science industries.

Cast, which has raised $18 million in venture capital to date, has a workforce of 52 employees and plans to have at least 80 employees by the end of the year.

“Cloud resources should come from any provider, and customers should leverage the best cost and performance resources available in real time,” Frayman said. “This cannot be a human-driven decision in order to scale.”

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

OpenAI releases Triton, a programming language for AI workload optimization

All the sessions from Transform 2021 are available on-demand now. Watch now.


OpenAI today released Triton, an open source, Python-like programming language that enables researchers to write highly efficient GPU code for AI workloads. Triton makes it possible to reach peak hardware performance with relatively little effort, OpenAI claims, producing code on par with what an expert could achieve in as few as 25 lines.

Deep neural networks have emerged as an important type of AI model, capable of achieving state-of-the-art performance across natural language processing, computer vision, and other domains. The strength of these models lies in their hierarchical structure, which generates a large amount of highly parallelizable work well-suited for multicore hardware like GPUs. Frameworks for general-purpose GPU computing such as CUDA and OpenCL have made the development of high-performance programs easier in recent years. Yet, GPUs remain especially challenging to optimize, in part because their architectures rapidly evolve.

Domain-specific languages and compilers have emerged to address the problem, but these systems tend to be less flexible and slower than the best handwritten compute kernels available in libraries like cuBLAS, cuDNN or TensorRT. Reasoning about all these factors can be challenging even for seasoned programmers. The purpose of Triton, then, is to automate these optimizations, so that developers can focus on the high-level logic of their code.

“Novel research ideas in the field of deep learning are generally implemented using a combination of native framework operators … [W]riting specialized GPU kernels [can improve performance,] but [is often] surprisingly difficult due to the many intricacies of GPU programming. And although a variety of systems have recently emerged to make this process easier, we have found them to be either too verbose, lack flexibility, generate code noticeably slower than our hand-tuned baselines,” Philippe Tillet, Triton’s original creator, who now works at OpenAI as a member of the technical staff, wrote in a blog post. “Our researchers have already used [Triton] to produce kernels that are up to 2 times more efficient than equivalent Torch implementations, and we’re excited to work with the community to make GPU programming more accessible to everyone.”

Simplifying code

According to OpenAI, Triton — which has its origins in a 2019 paper submitted to the International Workshop on Machine Learning and Programming Languages — simplifies the development of specialized kernels that can be much faster than those in general-purpose libraries. Its compiler simiplifies code and automatically optimizes and parallelizes it, converting it into code for execution on recent Nvidia GPUs. (CPUs and AMD GPUs and platforms other than Linux aren’t currently supported.)

“The main challenge posed by our proposed paradigm is that of work scheduling — i.e., how the work done by each program instance should be partitioned for efficient execution on modern GPUs,” Tillet explains in Triton’s documentation website. “To address this issue, the Triton compiler makes heavy use of block-level data-flow analysis, a technique for scheduling iteration blocks statically based on the control- and data-flow structure of the target program. The resulting system actually works surprisingly well: our compiler manages to apply a broad range of interesting optimization automatically.”

The first stable version of Triton, along with tutorials, is available from the project’s GitHub repository.

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

Database optimization startup Silk raises $55M

Join AI & data leaders at Transform 2021 on July 12th for the AI/ML Automation Technology Summit. Register today.


Silk, a platform designed to improve database performance, today announced it has raised $55 million in a series B funding round led by S Capital. The company says the funds will be used to support its sales and marketing operations and expand its engineering team as the demand for cloud environments rises in the wake of the pandemic.

As companies increasingly move their workloads to the cloud, stable cloud infrastructure is essential. Seventy percent of companies now use more than one database in their stack, according to Redgate. And in a recent Gleanster survey, respondents said databases are the No. 1 source of app performance issues.

Silk, which was founded in 2008 by CEO Dani Golan, Moshe Selfin, and Ofir Dubovi, enables customers to scale databases on-demand while delivering ostensibly better performance compared with conventional setups. The platform automatically detects and removes duplicate data, mitigating gaps between data and applying compression to bridge silos between clouds and on-premise servers.

Golan claims the platform makes cloud environments run up to 10 times faster and the entire app stack more resilient to malfunctions. “The cloud vendors are now beginning the fight over customers’ databases and other mission-critical ‘crown jewels.’ To win this fight, they need to guarantee that customers will meet their own end users’ service-level agreements by enabling prime scalability and performance of their mission-critical applications,” he added. “Having this capital allows us to bring the vision-accelerated cloud adoption to a wider audience.”

Database optimization

Silk, which sits between cloud environments and a customer’s databases, allocates cloud capacity to deliver automatic provisioning and clones data for free, letting companies create test environments without spending or slowing workload. The platform’s machine learning-powered dashboard shows real-time metrics and allows admins to scale up or down on the fly, as well as viewing the features and data they’re using.

Cloud observability and management remains a major challenge in the enterprise. In a Pepperdata report,  64% of respondents said cost management and containment was their biggest concern when it came to running cloud big data technologies and applications. Moreover, a majority of respondents said optimizing current cloud resources was their highest-priority big data cloud initiative.

Silk counts Payoneer, Priceline, Cisco, and Telefonica among its customers and has partnered with Microsoft, Amazon Web Services, and Google Cloud Platform to sell its technology and services. To date, the Needham, Massachusetts-based company has raised over $294 million in venture capital, including a $75 million series F round led by CIRTech Fund and Waterwood Group that closed in January 2017.

The most recent funding round included participation from Sequoia Capital, Pitango, Globespan, Ibex, Vintage, Clal Insurance, Bank HaPoalim, Meitav Dash, and Menora Mivtachim.

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

Adapdix adds adaptive AI tool EdgeOps DataMesh for process optimization %

Elevate your enterprise data technology and strategy at Transform 2021.


Adapdix on Thursday unveiled EdgeOps DataMesh, adaptive AI software that ingests, pre-processes and performs AI-driven analysis of data at the network edge to make automated, real time manufacturing process improvements.

DataMesh is intended to be especially useful in large-scale manufacturing. A pair of leading semiconductor manufacturers are among the first customers to incorporate the new product, which promises to streamline operations in a field that has recently been hit hard by supply-chain and production challenges.

The DataMesh software provides manufacturers with the ability to leverage data generated at the edge by “stitching together disparate data streams [for] edge inferencing in milliseconds,” Adapdix CEO Anthony Hill said in a statement. This leads to “real-time analysis, enabling split-second decision-making for critical operations and reduced downtime of high-value assets.”

Adapdix is based in Pleasanton, Calif. and was founded in 2015. The software developer’s edge-optimized data management platform speeds AI and machine learning to provide improvements to manufacturing services and processes. DataMesh is the first product to be released as part of the company’s EdgeOps platform.

EdgeOps improves data processing 

Hill said that, because the Adapdix platform operates at the source on the equipment, it benefits from ultra-low latency required for integrating advanced AI applications.

“With the ongoing worldwide chip shortage, semiconductor companies are urgently looking for ways to yield more from their existing resources and Adapdix software helps them achieve exactly that. Chip makers using DataMesh are better able to optimize the performance and efficiency of their manufacturing equipment to maximize the yield, throughput and quality of their semiconductor manufacturing processes,” he said.

Notably, Micron Technology is both an investor in Adapdix and a customer. The Boise, Idaho-based flash memory and storage manufacturer’s Micron Ventures was part of a recent funding round for Adapdix that also included WRVI Capital, SoftBank’s Opportunity Fund and X2 Equity.

“Adapdix provides an innovative platform that allows enterprises to operationalize artificial intelligence across critical infrastructure to unlock new levels of productivity and streamlined processes,” Micron director of venture capital Andrew Byrnes said. He indicated such software is a step toward fulfilling the promise of Industry 4.0.

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

Formation launches AI-driven sales offer optimization platform

Join Transform 2021 this July 12-16. Register for the AI event of the year.


Marketers send out billions of offers each year to engage customers. But succes depends on creating the optimal offer for each customer, a challenge that’s been compounded by changes in customer behaviors during the pandemic. A recent Forrester study found that only 33% of customers believe offers they received from brands were relevant.

Marketing tech startup Formation claims it has a solution in Dynamic Offer Optimization, a platform that enables travel, retail, and quick service restaurants to create, deploy, and optimize sales offers. Formation claims Dynamic Offer Optimization can increase customer engagement by helping businesses respond more quickly to market forces and changing customer needs.

Under the hood

Dynamic Offer Optimization consists of two components. There’s Offer Builder, which creates and optimizes individual offers, and Offer Engine, which drives real-time delivery, tracking, fulfillment, and measurement of each offer. Offer Builder ingests segments and customer journey data from marketing tech stacks and builds millions of unique offer variants. Offer Engine makes the offers available via APIs to channels and tracks, fulfills, and measures each offer’s performance.

Formation applies machine learning to optimize each customer’s offer for subsequent deployments. The company says this process can take less than an hour and that customers — including Starbucks and United Airlines — have used it to deliver over 2 billion unique offers during the pandemic.

“We’re really excited to take all the knowledge and experience from working with some of the world’s biggest brands for the past five years and make that breakthrough technology available to all companies,” Formation cofounder and CEO Christian Selchau-Hansen said in a press release. “Automating the creation, deployment, and optimization of billions of offers not only drives material impact to the business but enables brands and marketers to also focus on all the other critical mindset and organizational shifts needed for comprehensive digital transformation.”

New community

The release of Dynamic Offer Optimization coincides with the launch of Loyalty Innovators. As Selchau-Hansen explains, Loyalty Innovators’ mission is to connect and support digital, marketing, and loyalty leaders in their journeys to adapt their company’s products to the changing consumer landscape.

“We are excited to offer marketers unique technology to help them engage customers with relevant and valuable offers with tremendous speed and agility by better leveraging first-party data through automation and machine learning,” Selchau-Hansen continued. “Our mission is to arm digital and marketing leaders with best-in-class optimization tech, as well as to support them through their digital transformation journey with the Loyalty Innovators community.”

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

OctoML raises $28M for machine learning deployment optimization

Join Transform 2021 for the most important themes in enterprise AI & Data. Learn more.


Studies like the 2020 State of AI report from McKinsey have found that businesses capable of deploying multiple AI models are considered high performers, but a survey of business leaders included in the report found fewer than 20% have taken deep learning projects beyond the pilot stage. It’s well known that most businesses face challenges deploying AI in production, which has led to a rise in startups that serve needs like AIOps or auditing. In the latest news for such a company, OctoML today raised a $28 million series B funding round.

OctoML helps businesses accelerate AI model inference and training and relies on the open source Apache TVM machine learning compiler framework. TVM is currently being used by companies like Amazon, AMD, Arm, Facebook, Intel, Microsoft, and Qualcomm. OctoML will use the funding to continue building out products like its Octomizer platform and investing in its go-to-market strategy and customer service teams.

“We started the TVM work as a research project at the University of Washington about five years ago, and all the key people in the project are part of — they all got their Ph.D.s and are part of the company now,” OctoML CEO and cofounder Luis Ceze told VentureBeat. “We’re focused on making inference fast on any hardware, and support cloud and edge deployments.” 

Last month, OctoML joined more than 20 startups — including Algorithmia and Determined AI — that have banded together to create the AI Infrastructure Alliance, an effort to promote interoperability between the offerings from AI startups and advance alternatives to popular cloud AI services.

The $28 million funding round was led by Addition Capital, with participation from existing investors Madrona Venture Group and Amplify Partners.

OctoML has raised $47 million to date, including a $3.9 million seed funding round in October 2019, just months after the company was founded. OctoML is based in Seattle with remote employees across the United States.

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