Nvidia launches new services for training large language models

As interest around large AI models — particularly large language models (LLMs) like OpenAI’s GPT-3 — grows, Nvidia is looking to cash in with new fully managed, cloud-powered services geared toward enterprise software developers. Today at the company’s fall 2022 GTC conference, Nvidia announced the NeMo LLM Service and BioNeMo LLM Service, which ostensibly make it easier to adapt LLMs and deploy AI-powered apps for a range of use cases including text generation and summarization, protein structure prediction and more.

The new offerings are a part of Nvidia’s NeMo, an open source toolkit for conversational AI, and they’re designed to minimize — or even eliminate — the need for developers to build LLMs from scratch. LLMs are frequently expensive to develop and train, with one recent model — Google’s PaLM — costing an estimated $9 million to $23 million leveraging publicly available cloud computing resources.

Using the NeMo LLM Service, developers can create models ranging in size from 3 billion to 530 billion parameters with custom data in minutes to hours, Nvidia claims. (Parameters are the parts of the model learned from historical training data — in other words, the variables that inform the model’s predictions, like the text it generates.) Models can be customized using a technique called prompt learning, which Nvidia says allows developers to tailor models trained with billions of data points for particular, industry-specific applications — e.g. a customer service chatbot — using a few hundred examples.

Developers can customize models for multiple use cases in a no-code “playground” environment, which also offers features for experimentation. Once ready to deploy, the tuned models can run on cloud instances, on-premises systems or through an API.

The BioNeMo LLM Service is similar to the LLM Service, but with tweaks for life sciences customers. Part of Nvidia’s Clara Discovery platform and soon available in early access on Nvidia GPU Cloud, it includes two language models for chemistry and biology applications as well as support for protein, DNA and chemistry data, Nvidia says.

Nvidia LLMs

Visualization of bio processes predicted by AI models.

BioNeMo LLM will include four pretrained language models to start, including a model from Meta’s AI R&D division, Meta AI Labs, that processes amino acid sequences to generate representations that can be used to predict protein properties and functions. Nvidia says that in the future, researchers using the BioNeMo LLM Service will be able to customize the LLMs for higher accuracy

Recent research has shown that LLMs are remarkably good at predicting certain biological processes. That’s because structures like proteins can be modeled as a sort of language — one with a dictionary (amino acids) strung together to form a sentence (protein). For example, Salesforce’s R&D division several years ago created an LLM model called ProGen that can generate structurally, functionally viable sequences of proteins.

Both the BioNeMo LLM Service and LLM Service include the option to use ready-made and custom models through a cloud API. Usage of the services also grants customers access to the NeMo Megatron framework, now in open beta, which allows developers to build a range of multilingual LLM models including GPT-3-type language models.

Nvidia says that automotive, computing, education, healthcare and telecommunications brands are currently using NeMo Megatron to launch AI-powered services in Chinese, English, Korean and Swedish.

The NeMo LLM and BioNeMo services and cloud APIs are expected to be available in early access starting next month. As for the NeMo Megatron framework, developers can try it via Nvidia’s LaunchPad piloting platform at no charge.

Nvidia launches new services for training large language models by Kyle Wiggers originally published on TechCrunch

Nvidia debuts new products for robotics developers, including Jetson Orin Nano

Amid the festivities at its fall 2022 GTC conference, Nvidia took the wraps off new robotics-related hardware and services aimed at companies developing and testing machines across industries like manufacturing. Isaac Sim, Nvidia’s robotics simulation platform, will soon be available in the cloud, the company said. And Nvidia’s Jetson lineup of system-on-modules is expanding with Jetson Orin Nano, a system designed for low-powered robots.

Isaac Sim, which launched in open beta last June, allows designers to simulate robots interacting with mockups of the real world (think digital re-creations of warehouses and factory floors). Users can generate data sets from simulated sensors to train the models on real-world robots, leveraging synthetic data from batches of parallel, unique simulations to improve the model’s performance.

It’s not just marketing bluster, necessarily. Some research suggests that synthetic data has the potential to address many of the development challenges plaguing companies attempting to operationalize AI. MIT researchers recently found a way to classify images using synthetic data, and nearly every major autonomous vehicle company uses simulation data to supplement the real-world data they collect from cars on the road.

Nvidia says that the upcoming release of Isaac Sim — which is available on AWS RoboMaker and Nvidia NGC, from which it can be deployed to any public cloud, and soon on Nvidia’s Omniverse Cloud platform — will include the company’s real-time fleet task assignment and route-planning engine, Nvidia cuOpt, for optimizing robot path planning.

“With Isaac Sim in the cloud … teams can be located across the globe while sharing a virtual world in which to simulate and train robots,” Nvidia senior product marketing manager Gerard Andrews wrote in a blog post. “Running Isaac Sim in the cloud means that developers will no longer be tied to a powerful workstation to run simulations. Any device will be able to set up, manage and review the results of simulations.”

Jetson Orin Nano

Back in March, Nvidia introduced Jetson Orin, the next generation of the company’s Arm-based, single-board PCs for edge computing use cases. The first in the line was the Jetson AGX Orin, and Orin Nano expands the portfolio with more affordable configurations.

Nvidia Jetson Orin Nano

Image Credits: Nvidia

The aforementioned Orin Nano delivers up to 40 trillion operations per second (TOPS) — the number of computing operations the chip can handle at 100% utilization — in the smallest Jetson form factor to date. It sits on the entry-level side of the Jetson family, which now includes six Orin-based production modules intended for a range of robotics and local, offline computing applications.

Coming in modules compatible with Nvidia’s previously announced Orin NX, the Orin Nano supports AI application pipelines with Ampere architecture GPU — Ampere being the GPU architecture that Nvidia launched in 2020. Two versions will be available in January starting at $199: The Orin Nano 8GB, which delivers up to 40 TOPS with power configurable from 7W to 15W, and the Orin Nano 4GB, which reaches up to 20 TOPS with power options as low as 5W to 10W.

“Over 1,000 customers and 150 partners have embraced Jetson AGX Orin since Nvidia announced its availability just six months ago, and Orin Nano will significantly expand this adoption,” Nvidia VP of embedded and edge computing Deepu Talla said in a statement. (By comparison to the Orin Nano, the Jetson AGX Orin costs well over a thousand dollars — needless to say, a substantial delta.) “With an orders-of-magnitude increase in performance for millions of edge AI and [robotics] developers Jetson Orin Nano sets new standard for entry-level edge AI and robotics.”

Nvidia debuts new products for robotics developers, including Jetson Orin Nano by Kyle Wiggers originally published on TechCrunch

Salesforce, Snowflake partnership moves customer data in real time across systems

Salesforce is the best selling CRM in the world, and Snowflake is one the top cloud data lakes. The latter lets customers store and manage massive amounts of unstructured data. When you mix the two services, it has the potential to be a powerful combination.

The two companies have been working together for some time, but ahead of the Dreamforce customer conference in San Francisco next week, they announced an enhancement to that partnership where data can flow freely between the Snowflake data repository and the Salesforce customer data platform (CDP).

The idea, says David Schmaier, president and chief product officer at Salesforce is to provide, to the extent possible, a single, up-to-date customer record in real time with the ultimate goal of optimizing the customer experience based on what the company knows about you.

He says that Salesforce starts with the core belief that companies with the best data can build the smartest machine learning models. “If we can enrich and unify and deepen the data, then your AI can do more, and if your AI can do more, then your customer interactions are that much more tailored and personalized,” Schmaier explained.

But to get to that point requires a CDP, a tool that collects all the data about a customer’s interactions with a company in one central repository. The CDP operates best with real-time data, and Snowflake can be the source of that data. It helps that Schmaier says the company’s CDP customers tend to use both tools, making the partnership even more valuable for both companies.

Christian Kleinerman, SVP of product at Snowflake, says that while the relationship goes back long before this year, this level of integration is new. “[We talked about] how we could bring Salesforce CRM system data onto the Snowflake data cloud, then let customers create interesting solutions, interesting outcomes, but also feed that data back into Salesforce itself – and that is at the heart of the integration,” Kleinerman said.

The reality of integrating data across systems is rather daunting. Schmaier points out that customers often have hundreds or even thousands of data sources connected to the CDP, and when you think about the amount of data moving through Snowflake on top of that, it’s a tremendous amount of information they have to process to make this work.

While creating the best customer experience is the goal, the two companies realize this is the ideal, and as companies work to understand and process the data, it brings them closer to building personalized customer experiences at scale, which remains the holy grail of online sales.

One of the advantages of working with Snowflake is the notion of “zero copies,” which means with all this data floating around, they don’t have to make copies of it to make this work. Instead, Kleinerman says the technology takes advantage of references to point to the data.

“So instead of copying CDP data onto our mutual customers’ Snowflake account, what Salesforce does is it leverages that data sharing technology to make the CDP data available for querying on the Snowflake side of our mutual customers. So now they can join it, enrich it or run it through machine learning. But if the data changes in Salesforce in the CDP, it is reflected in Snowflake in real time,” Kleinerman said.

Like many things being announced at (or ahead of) Dreamforce, this is not yet available, but will be in closed pilot starting this Fall.

Salesforce, Snowflake partnership moves customer data in real time across systems by Ron Miller originally published on TechCrunch

Bluesky built cost guardrails to help cut Snowflake data spend

Snowflake has a revenue model that investors have to love, but big customers, not so much. That’s because it’s based on a consumption model where the more you use, the more you pay — and when it comes to data management these days, that can add up pretty quickly.

Bluesky, a new startup from a couple of ex-Google and Uber engineers, came up with a way to help reduce those bills, and today the company announced a healthy $8.8 million seed.

Mingsheng Hong, Bluesky co-founder and CEO, who spent more than eight years at Google, says that Bluesky takes an organized approach to cost cutting. “First, we observe to get the visibility to understand who has been spending and what the most expensive workloads and queries are,” Hong explained. He jokingly referred to this step as “the walk of shame because finally everyone knows how much you’re spending.”

The idea is to delete workloads that are costing cash, but don’t add a lot of value.

The second piece involves optimizing the remaining spend by finding what the most expensive workloads are and figuring out how to adjust them to reduce the overall cost. He said it’s often about simply changing how they run the query, taking a smarter approach to save money.

“These queries will still run, and they will still get the same result, but we run them faster and cheaper by, for example, avoiding scanning a huge table, when we can generate indices [to get the same result] without having to scan a very large data set.”

The last step, which is planned for the future, is to have an optimization engine that automatically does this for you. Once customers trust the software to do the job, it will constantly be scanning the workloads and searching for ways to cut costs automatically.

You may think that Snowflake would be threatened by such a product, but Hong says the company is actually a partner. “Snowflake is bringing us in to help customers to reduce or manage growth, and this way all three parties are happy. Customers can sign a larger contract with Snowflake, and yet they know they have the assurance that they have the cost guardrails that Bluesky provides,” he said.

He said the product is built around optimizing SQL queries, and that it plans to apply the same approach to other consumption-based data products like Databricks in the future.

The company is just six months old, but has more than 10 customers using the product including Coinbase. Hong launched the company with CTO Zheng Shao, who helped build the open source project Apache Hive, a SQL query engine built on top of Hadoop, an early way of dealing with large data sets.

The company currently has 15 employees, and he is trying to build a diverse group right out of the gate. “In terms of diversity, we make sure that we have people coming in with different perspectives. They may have a different perspective due to their culture or gender, but first and foremost, it’s not about trying to diversify for the sake of diversity, hitting some metrics. It’s about people coming in with different ideas and perspectives,” Hong said.

Today’s $8.8 million seed investment was led by Greylock with participation from several industry angels.

Bluesky built cost guardrails to help cut Snowflake data spend by Ron Miller originally published on TechCrunch

Zesty lands $75M for tech that adjusts cloud usage to save money

Spending on the cloud shows no signs of slowing down. In the first quarter of 2021, corporate cloud services infrastructure investment increased to $41.8 billion, representing 35% year-on-year growth, according to Grand View Research. But while both small- and medium-sized businesses and enterprises admit that they’re spending more on the cloud, they’re also struggling to keep costs under control. According to a 2020 Statista survey, companies estimate that 30% of their cloud spend is ultimately wasted.

The desire to better manage cloud costs has spawned a cottage industry of vendors selling services that putatively reign in companies’ infrastructure spending. The category grows by the hour, but one of the more successful providers to date is Zesty, which automatically scales resources to meet app demands in real time.

Zesty today closed a $75 million Series B round co-led by B Capital and Sapphire Ventures with participation from Next47 and S Capital. Bringing the company’s total raised to $116 million, the proceeds will be put toward supporting product development and expanding Zesty’s workforce from 120 employees to 160 by the end of the year, CEO Maxim Melamedov tells TechCrunch.

“DevOps engineers … face limitations such as discount program commitments and preset storage volume capacity, CPU and RAM, all of which cannot be continuously adjusted to suit changing demand,” Melamedov said in an email interview. “This results in countless wasted engineering hours attempting to predict and manually adjust cloud infrastructure as well as billions of dollars thrown away each year.”

Melamedov founded Zesty with Alexey Baikov in 2019, after the pair observed that cloud infrastructure wasn’t keeping up with the pace of change in business environments. Prior to co-launching Zesty, Melamedov was the VP of customer success at Gimmonix, a travel tech company. He briefly worked together with Baikov at big data firm Feedvisor. Baikov was previously a DevOps team lead at Netvertise.


Image Credits: Zesty

At the core of Zesty is an AI model trained on real-world and “synthetic” cloud resource usage data that attempts to predict how many cloud resources (e.g., CPU cores, hard drives and so on) an app needs at any given time. The platform takes actions informed by the model’s projections, like automatically shrinking, expanding and adjusting storage volume types and purchasing and selling public cloud instances.

To increase or decrease storage, Zesty transforms filesystem volumes in the cloud into a virtual disk with a series of multiple volumes, each of which can be expanded or shrunk. On the compute side, the platform collects real-time performance metrics, buying or selling cloud compute in response to app usage.

“The primary tools we use to design efficient automation of cloud resources come from the fields of decision analysis and resource management. Many of the classical techniques used to solve such problems can be slow and not suitable for real-time decision making, where fast response to change is critical,” Melamedov said. “With Zesty, organizations dramatically reduce cloud costs and alleviate the burdensome task of managing cloud resources in a constantly shifting business environment. Because in a world that’s always changing, Zesty enables the infrastructure to change right around with it.”

Those are lofty promises to be sure. But Zesty has managed to grow its customer base to over 300 companies, including startups Heap, Armis and WalkMe, suggesting that it’s doing something right.

[T]he pandemic create[d] a whole new level of demand for our solutions and we have been fortunate to see huge demand growth for our products,” Melamedov said. “Companies were not only looking to save money, but they were [also] forced to cut staff. Freeing up DevOps and other operational personnel became critically important, and that’s where we came in — freeing them up from having to babysit the cloud and constantly be on call to adjust cloud resources as needs shifted. The current [economic] slowdown as well has only helped showcase our value even more, now that we have dozens of case studies we can share that show quick and easy return on investment.

Zesty’s challenge will be continuing to stand out in a field of rivals. Microsoft in 2017 acquired Cloudyn, which provided tools to analyze and forecast cloud spending. Then, in 2019, Apptio snatched up cloud spending management vendor Cloudability, while VMware, NetApp and Intel bought CloudHealth, Spot (formerly Spotinst) and Granulate, respectively, within the span of a few years. Elsewhere, ventures such as Granulate, Cast AI, Exotanium and Sync Computing have raised tens of millions of venture capital dollars for their cloud spend-optimizing tech.

Melamedov wouldn’t go into specifics around Zesty’s financials. But he expressed confidence in the company’s prospects, revealing that Zesty has reached an annual run rate in the “tens of millions.”

Zesty lands $75M for tech that adjusts cloud usage to save money by Kyle Wiggers originally published on TechCrunch

Opus Security emerges from stealth to help tackle cloud security threats

Opus Security, a cloud security orchestration and remediation platform, today emerged from stealth with $10 million in seed funding led by YL Ventures, with participation from Tiger Global and angel investors. CEO Meny Har tells TechCrunch that the proceeds will be put toward launching Opus’ platform in general availability, expanding the startup’s footprint in the U.S. and product R&D.

It’s Har’s assertion that cloud security teams rely heavily on manual processes to resolve security incidents, which isn’t scalable. He’s not an unbiased source, exactly. But to his point, a survey commissioned by Orca Security found that 59% of security teams receive more than 500 alerts about public cloud security every day. In a separate poll from ISACA and HCL Technologies released last May, 61% of IT security professionals said their teams were understaffed.

“Today, in order to remediate the multiplying number of security findings, it has become necessary to include a wide array of teams and stakeholders within the organization in security processes critical for remediation,” Har told TechCrunch in an email interview. “This complex collaboration — with no streamlined orchestration process — creates friction and time drainage as the teams fail to communicate clearly. This friction leads to a general lack of visibility into how well the organization is secure, how remediation processes are undertaken and what needs to be improved.”

Har co-founded Opus with Or Gabay, with whom he worked at security orchestration startup Siemplify. Har was a member of the founding team at Siemplify, which was acquired by Google earlier this year.

“We witnessed firsthand the challenges security operations teams face when trying to analyze, prioritize and remediate security risks using cumbersome, distributed processes and various detection tools — without oversight or management,” Har said. “Opus’ vision is to empower security operations teams to see beyond alerts and threats and gain knowledge, capabilities and control to dramatically cut down the time to resolve them.”

To this end, Opus draws on operational and technical data from existing cloud security tools to create a “connective tissue” between security operations teams and other enterprise departments. The attempts to orchestrate the response and remediation process with guidelines and playbooks, leveraging automation to resolve issues that commonly don’t need human involvement — and delivering key metrics along the way.

“Organizations today may choose to use a common ticketing platform such as ServiceNow or Jira and potentially add ‘background’ security automation platforms, used to craft processes manually to support the security operations and DevOps functions. Generic ticketing platforms may help with management but do little in terms of efficiency through automation,” Har said. “With instant visibility and mapping of remediation … Opus removes blind spots and provides security and business executives with immediate and tangible insights into the state of their risk.”

That’s a lot to promise — especially in the face of competition like cloud security startups Wiz, Paladin Cloud and Laminar. But in an encouraging sign (potentially), Har says that Opus has seen early adoption among a “handful” of design partners, who are working to build out the platform ahead of a broad launch sometime in Q1 2023. In the lead-up to general availability, Opus plans to expand its headcount from 10 employees currently to 20 to 25 by the end of the year, according to Har.

“As organizations recover from the aftermath of the COVID-19 pandemic, and with looming budget and workforce cuts plaguing the industry, Opus is a tailored solution devised specifically with these external effects in mind,” Har said. “Implementing Opus’ solution is simple, and as Opus drives automation within all remediation processes, it is the right fit for organizations striving to do more with less.”

Like many other startups, Opus is benefiting from VC dollars that, despite the macroeconomic downturn, haven’t stopped flowing in the cybersecurity sector. According to Momentum Cyber’s latest cybersecurity market review, investors poured $11.5 billion in total venture capital financing into cybersecurity startups in the first half of 2021, up from $4.7 billion during the same period a year earlier.

John Brennan, senior partner at YL Ventures, added in an emailed statement: “The proliferation of cloud-focused security solutions has dramatically raised organizational awareness to the scope of their risk surface. While visibility in the cloud has greatly improved, customers now express a need for a dedicated solution to address the drastically increasing number of alerts … Meny and Or have leveraged their unique experience at Siemplify to build the industry’s first cloud-native
remediation orchestration and automation platform.”

Opus Security emerges from stealth to help tackle cloud security threats by Kyle Wiggers originally published on TechCrunch

Microsoft launches Arm-based Azure VMs powered by Ampere chips

Following a preview in April, Microsoft this morning announced the general availability of virtual machines (VMs) on Azure featuring the Ampere Altra, a processor based on the Arm architecture. The first Azure VMs powered by Arm chips, Microsoft says that they’re accessible in ten Azure regions today and can be included in Kubernetes clusters managed using Azure Kubernetes Service beginning on September 1.

The Azure Arm-based VMs have up to 64 virtual CPU cores, 8GB of memory per core and 40Gbps of networking bandwidth as well as SSD local and attachable storage. Microsoft describes them as “engineered to efficiently run scale-out, cloud-native workloads,” including open source databases, Java and .NET applications and gaming, web, app and media servers.

Preview releases of Windows 11 Pro and Enterprise and Linux OS distributions including Canonical Ubuntu, Red Hat Enterprise Linux, SUSE Enterprise Linux, CentOS and Debian are available on the VMs day one, with support for Alma Linux and Rocky Linux to arrive in the future. Microsoft notes that Java apps in particular can run with few additional code changes, thanks to the company’s contributions to the OpenJDK project.

The launch of the Azure VMs is a notable win for Ampere, which came out of stealth in 2018 with the ambitious goal of competing with Intel for a slice of the ~$10 billion data center chip market. Backed by $426 million in venture capital and led by a former Intel president, the company has managed to snag a foothold in recent years, inking deals with Oracle, Equinix, Google Cloud and China-based cloud service providers Tencent Cloud, JD Cloud and UCloud to launch Arm-based VMs.

Ampere competes with Arm-powered VMs from Amazon Web Services, which acquired startup Annapurna Labs in 2015 to build its own Arm-based, general-purpose server hardware lineup called Graviton. Microsoft is reportedly pursuing its own Arm chip designs, as well, as are Chinese tech giants Alibaba and Huawei.

Research firm Omdia said last August that it expects Arm to account for 14% of servers by 2025. If the prediction comes to pass, it’d be a major coup against Intel’s x86 chips, which controlled an estimated 89% of the market as of March 2022.

For Microsoft, the Ampere VMs launch is a step toward fulfilling the pledge it made five years ago to power more than half of its cloud data center capacity with Arm-compatible servers. After a false start with Centriq server processors from Qualcomm, which were ultimately discontinued, the company appears better-positioned to reach that threshold.

“The general availability of Microsoft Azure VMs on Arm marks an important milestone in redefining what is possible in cloud computing,” Arm SVP Chris Bergey is quoted as saying in a blog post detailing the Azure VMs. “Through market-leading scalable efficiency and the liberty to innovate, Arm … is enabling Azure customers to embrace the increasing diversity of workloads with better overall total cost of ownership and cleaner cloud service operations.”

Rookout raises $16M Series B to scale its developer-first observability platform

Rookout, the Tel Aviv-based startup that describes itself as a ‘developer-first observability platform,’ today announced that it has raised a $16 million Series B funding round led by Fort Ross Ventures. Existing investors TLV Partners, Emerge and Cisco Investments, as well as new investors LIAN Group, Mighty Capital and Binder & Partners, also participated in this round, which brings the company’s total funding to over $28 million.

The promise of Rookout is to give engineers more data about how their code runs in production. That, the company argues, sets it apart from more traditional monitoring tools which tend to focus more on the infrastructure and helping SREs do their job, and not the live code and business logic that developers care about.

Image Credits: Rookout

“We’re trying to give developers ownership over production — because that’s what we care about,” Rookout CEO Liran Haimovitch told me. “At Rookout, with 20 engineers, I don’t care what they’re doing on their laptops. I honestly don’t care. I care about what their code is doing in production. I think every engineering leader out there is feeling the same. But traditionally, engineers didn’t have access to what’s going on in production, and you can’t really make people care about something if you’re keeping them away from it.”

The company’s live debugging features are at the core of its toolset, powered by its dynamic instrumentation capabilities that enable developers to set what the company calls ‘non-breaking breakpoints’ to collect live data and debug their applications in production. But the tool also integrates data from tracing tools like OpenTracing and OpenTelemtry, as well as various other third-party logging services.

The company says its customer base now includes the likes of Amdocs, Cisco, Dynatrace, Jobvite, Santander Bank and UPS. Since its last funding round in 2019, Rookout saw its revenue increase by 20x.

“We’ve been impressed with Rookout’s execution of its groundbreaking solution, alongside the rapid trajectory of its growing customer base and significant expansion momentum within the enterprise,” said Sharin Fisher Dibrov, partner at Fort Ross Ventures. “We are coming to a third wave of observability tools which shifts everything further left, and we’re excited to support Rookout’s journey and backing [Rookout CEO] Shahar [Fogel], Liran, and the team as the category-defining leader in developer-first observability.”

Omni looks to take on Looker with its cloud-powered BI platform

There’s been an explosion of business intelligence (BI) tools in recent years, or tools that analyze and convert raw data into info for use in decision making. Investments in them are on the rise, but companies are still struggling to become “data-driven” — at least, according to some survey results. NewVantage Partners’ 2022 poll of chief data and analytics officers found that less than half (47.4%) believed that they’re competing on data and analytics. They cited company culture and the overwhelming growth of data as the top blockers, as well as concerns over data ownership and privacy.

Colin Zima believes that there’s another major challenge businesses adopting BI tools have to overcome: poor usability. He’s the co-founder and CEO of Omni, a BI platform that aims to simplify working with data across an organization. As such, Zima might not be incredibly impartial. But on the other hand, he’s a longtime participant in the data analytics community, having worked at Google on the Search quality team and at Looker as the chief analytics officer and VP of product.  

“In an era where every employee is expected to be a data user, getting the basics done is still way too hard: Looking up data across many different systems or waiting on the data team to pull data or being forced to learn structured query language (SQL) to answer questions,” Zima said. “The reality is business users need great, simple tools to do their jobs better and data teams need powerful tools to manage that process and do high-value work that complements core reporting.”

Zima co-launched Omni in early 2022 alongside Jamie Davidson and Chris Merrick, who spent several years at Looker and Stitch, respectively, before joining the startup. The three co-founders were spurred by a mutual desire to build a product that made it easier for data teams to perform “high-value” work that complemented core business reporting processes, Zima said. 

“There are … some painful trade-offs folks make when they use a centralized platform — it feels so heavy to make changes, so folks were complementing with analyst tools or other point solutions and this fragmentation has only accelerated. This creates trade-offs — and really, tension — between data people and business teams, or folks that want to move quickly and your board reporting,” Zima said. “While legacy BI platforms unified teams around reliable, centralized data, it still meant a heavy upfront data modeling process. With Omni, we’re filling the gap between instant-gratification analytics and the reliability and governance of mature enterprise BI.”

Investors believe in Omni’s vision, having pledged $26.9 million toward the startup, including a seed round joined by Box Group, Quiet and Scribble and a $17.5 million Series A led by Redpoint with participation from First Round and GV. Omni’s post-money valuation stands close to $100 million, according to a source familiar with the matter. As for the proceeds, Zima said they’ll be put toward go-to-market efforts; he claims that Omni still hasn’t spent the seed. 

Omni is comparable to existing BI tools like the aforementioned Looker and Tableau, Zima says. But the platform can also take raw SQL — the language used to communicate with databases — and break it into modeled components. Omni’s built-in tools generate data models and components from SQL, creating a “sandbox” data model and allowing users to promote metrics to the official, shared model that the whole organization can use. Beyond this, Omni runs “automated aggregates” in-database to accelerate queries and manage costs for users (and their employers).

“The compromise most companies are forced to make with monolithic, centralized BI tools is that they hamstring employees and teams to work outside the core paths. That leaves the choice of either not using data or folding in shadow IT like Excel or isolated analytical tools to complete a workflow,” Zima said. To his point, research suggests that roughly half of organizations struggle to use and access quality data. “By bridging this gap between IT and business units, Omni is building a system that gives IT more control by promoting manageable decentralization versus just spinning up isolated tools to solve problems. Ultimately, this means all that business logic and data control can be retained and observed by IT and data teams, and thoughtfully integrated into core systems versus left on islands.

Launching a company during a downturn isn’t easy, although Zima says that Omni was insulated in many ways because of its founders’ longstanding relationships with Omni’s investors. Regardless of the macroeconomy, the core focuses this year will be hiring and customer acquisition, Zima says — Omni only worked with five development partners prior to today, which marks the platform’s public launch. Omni has about 16 employees currently and plans to expand that number by 25% by 2023.

The trick will be maintaining growth in the face of competition like Y42, Metabase and MachEye, the last of which raised $4.6 million in seed funding two years ago. More formidable is Pyramid Analytics, a business intelligence and analytics firm that landed $120 million last May. There’s also NoogataFractal AnalyticsTredence, LatentView and Mu Sigma.

For Zima’s part, he expects the down market to work in Omni’s favor at the expense of rivals as companies seek to consolidate their tools and “streamline their data stacks.”

“[Omni] is the only BI platform that combines the consistency of a shared data model with the freedom of SQL … [and] enables this virtuous feedback loop between one-off speed work and the governed model,” Zima said. “A core part of our thesis is that the central challenge that remains in business intelligence is tackling and uniting the entire surface area, which is mostly made up of point solutions … The emergence of the cloud data era [opens] up new, more ambitious possibilities like proactively optimizing performance.”

Sync Computing rakes in $15.5M to automatically optimize cloud resources

After a pandemic-driven cloud adoption boom in the enterprise, costs are finally coming under a microscope. More than a third of businesses report having cloud budget overruns of up to 40%, according to a recent poll by observability software vendor Pepperdata. A separate survey from Flexera found that optimizing the existing use of cloud services is a top initiative at 59% of companies — cost being the main motivation. 

An entire cottage industry of startups has sprung up around optimizing cloud compute. But one in the race, Sync Computing, claims to uniquely tie business objectives like cost and runtime reduction directly to low-level infrastructure configurations. Founded as a spinout from MIT’s Lincoln Laboratory, Sync today landed $12 million in a venture funding round (plus $3.5 million in debt) led by Costanoa Ventures, with participation from The Engine, Moore Strategic Ventures and National Grid Partners.

Sync co-founders Jeff Chou and Suraj Bramhavar both worked as members of the technical staff at the MIT Lincoln Laboratory prior to launching the startup. Bramhavar came to MIT by way of a photonics research position at Intel, while Chou co-founded another startup — Anoka Microsystems — designing a low-cost optical switch.

Sync was born out of innovations developed at the Lincoln Lab, including a method to accelerate a mathematical optimization problem commonly found in logistics applications. While many cloud cost solutions either provide recommendations for high-level optimization or support workflows that tune workloads, Sync goes deeper, Chou and Bramhavar say, with app-specific details and suggestions based on algorithms designed to “order” the appropriate resources.

“[We realized that our methods] can dramatically improve resource utilization of all large-scale computing systems,” Chou told TechCrunch in an email interview. “As Moore’s Law slows down, this will become a key technological choke point.”

Chou claims that Sync doesn’t require much in the way of historical data to begin optimizing data pipelines and provisioning low-level cloud resources. For example, he says, with just the data from a single previous run, some customers have accelerated their Apache Spark jobs by up to 80% — Apache Spark being the popular analytics source engine for data processing.

Sync recently released an API and “autotuner” for Spark on AWS EMR, Amazon’s cloud big data platform, and Databricks on AWS. Self-service support for Databricks on Azure is in the works.

“The launch of our public API will allow users to programmatically apply the Sync autotuner to a large number of jobs and enable continuous monitoring of [cloud environments] with custom integration,” Chou said. “The C-suite cares about managing cloud computing costs, and our Sync autotuner does this while also accelerating the output of data science and data analytics teams … The product also allows data engineers to quickly change infrastructure settings to achieve business goals. For example, one day, teams may need to minimize costs and de-prioritize runtime, but the next day, they may have a hard deadline, therefore needing to accelerate runtime. With Sync, this can be done with a single click.”

Sync first applied its technology inside MIT’s Supercomputing Center before working with larger government high-performance compute centers, including the Department of Defense — with which it has a $1 million contract. Now, Sync says it has roughly 300 registered users on its self-service app and “several dozen” design partners testing and providing feedback, including Duolingo and engineers at Disney’s Streaming Services group. 

“The pandemic and recent economic climate have been a boon for Sync, as controlling cloud costs through improved efficiency is now top of mind for many cloud software-as-a-service-native companies. Many companies are on hiring freezes and need an ‘easy button’ to drop cloud costs without adding burden or overhead to teams already at over capacity,” Chou said. “With the recent economic downturn, the demand for Sync’s unique approach has accelerated dramatically, already getting adopted by major enterprise customers. Our main challenge is for developers and CTOs to see how what we’ve built is different and also realize both can dramatically benefit by using it.”

Chou says the funding from the latest round, which bring’s Boston-based Sync’s total capital raised to $21.6 million, will be put toward customer acquisition, marketing and sales, product development, and R&D, including adding integrations with existing engineering workflows. Sync currently has 14 employees, a number that Chou expects will grow to 25 by the end of the year.