Michael Ovitz didn’t invent the idea of a talent agency, but one might argue that he perfected it. He founded the CAA in 1975, and grew it into the world’s leading talent agency, serving as chairman for 20 years. Now, Ovitz is investing in a brand new type of talent agency called Human Capital.

Human Capital is a hybrid organization, one part VC fund, one part recruiting business and one part creative agency. (Human Capital did not invest in its agency startup from its VC fund.) The Human Capital VC fund has $210 million in assets under management.

The Human Capital recruitment/agency company, founded by former General Catalyst associate Armaan Ali and Stanford grad Baris Akis, looks to provide for tech engineers the same services that Ovitz provided to actors and creatives back in the 70s, 80s and 90s. Engineers are some of the most sought-after talent in Silicon Valley and across the globe. And while big corporations and high-growth startups duke it out over these young engineers, the candidates themselves have little to no guidance around where they should go, what they should expect during the process, and, in some cases, what they should expect to earn.

Ovitz — alongside Qasar Younis, founder of Applied Intuition and former partner and COO of YC; Adam Zoia, founder and chairman of Glocap; Stephen Ehikian, co-founder and CEO of Airkit; and other financial institutions and LPs — recently injected $15 million into Human Capital, which is valued in the hundreds of millions according to the company.

Human Capital looks to pair the brightest engineers with the right company for them, while giving startups a new way to approach recruitment. Thus far, the company has 5,000 members (engineers) and has placed them at startups like Brex, Grammarly, Robinhood and more.

Human Capital starts by doing outreach on university campuses with outstanding engineering programs, setting up coffee with engineers who have been recommended or referred by alumni of the program. Once accepted as a member, the engineer explains to Human Capital what type of role they’re interested in, whether it’s at a big corporation, a high-growth startup or an early-stage company where they have the opportunity to build something from scratch.

The recruitment team at Human Capital then coaches the engineer through the interview process and beyond, helping with decision-making around promotions, understanding equity and negotiating new offers.

The org never charges the engineer, but rather takes a commission on the engineer’s annual income for the first year from the startup that recruited them.

Ali explained to TechCrunch how Human Capital is operating during the coronavirus pandemic, describing a situation in which the top talent that is in the market right now has a level of uncertainty about the future, leading them to seek positions at huge companies like Facebook and Google.

“Our hypothesis when we started this was that there are amazing businesses that are being run better at an earlier stage and have a proxy for that same type of stability [at a Google or Facebook] via their access to capital, alongside other foundational pieces of business security, such as their business model, unit economics, long-term vision for the company, gross margin rate, and growth opportunities for individuals at those companies.”

He said that Human Capital believed that, if a macro event occurred in the market place — we’re right in the middle of one of the least predictable and most impactful macro economic events ever — some of those “stable” earlier-stage businesses wouldn’t be hit in the same way as public companies who have to worry about short-term profitability.

“The issue is that you have to know a lot about those businesses in order to be able to discern that, and that’s our job,” said Ali. “And what we’ve seen is that a number of the companies in that position are actually ramping up recruiting right now.”

There is no mandatory link between Human Capital’s venture capital fund and their recruiting/agency entity, though the fund does like to invest in engineers who have gone through the program and move on to start their own businesses. Those types of investments include Brex, Bolt and Qualia, among others. Human Capital also invests in companies for whom they’ve recruited, such as Livongo, Snowflake, Clumio, Wildlife and Trackonomy. Human Capital has a preference for leading rounds only for companies that are started by its engineer members.

The model isn’t unlike SignalFire or Glocap, founded by Adam Zoia (investor in Human Capital). The idea is that VC funds are great for capital injections, but with the cut-throat recruiting atmosphere and a finite number of engineers, that money can be relatively useless if it can’t be used to bring on the best talent. So firms like SignalFire (in the tech world) and Glocap (in the business/finance world) put recruitment front and center in their value proposition. (Glocap doesn’t invest, but is the premier recruitment platform in the financial sector.)

Human Capital is also starting to look at potential acquisitions that can beef up its agency business, recently acqui-hiring Khonvo Corporation, a recruitment agency founded by Archit Bhise and Andrew Rising.

Ovitz explained to TechCrunch that his ultra-successful career as an agent stemmed from his ability to make decisions about people and projects quickly. He sees the same type of intuition in Ali and Akis at a much younger age and with less experience than he had.

“It’s a checklist in your head,” said Ovitz. “It’s a combination of when your brain meets your stomach, your intellect meets your gut that lets you know you’ve hit a winner. The thing that’s allowed Ali and Akis to build a company that’s worth the hundreds of millions in such a short period of time is that they had that when I met them without having an enormous amount of experience.”

He added that access to the internet, which he did not have during his agency days, is an amazing learning tool and an “epic crutch” that, when paired with good instincts, can accelerate the learning curve on building a business.

(It’s worth noting that this isn’t Ovitz’s first foray into Silicon Valley. The entertainment powerhouse was one of the earliest advisors to Marc Andreessen and Ben Horowitz during the formation of the legendary VC firm a16z, helping them model the firm after CAA itself. Ovitz has been quietly investing in and advising tech startups for the past 15 years.)

Google Meet, like all video chat products, is seeing rapid growth in user numbers right now, so it’s no surprise that Google is trying to capitalize on this and is quickly iterating on its product. Today, it is officially launching a set of new features that include a more Zoom-like tiled layout, a low-light mode for when you have to make calls at night and the ability to present a single Chrome tab instead of a specific window or your entire screen. Soon, Meet will also get built-in noise cancellation so nobody will hear your dog bark in the background.

If all of this sounds a bit familiar, it’s probably because G Suite exec Javier Soltero already talked to Reuters about these features last week. Google PR is usually pretty straightforward, but in this case, it moved in mysterious ways. Today, though, these features are actually starting to roll out to users, a Google spokesperson told me, and today’s announcement does actually provide more details about each of these features.

For the most part, what’s being announced here is obvious. The tiled layout allows web users to see up to 16 participants at once. Previously, that number was limited to four and Google promises it will offer additional layouts for larger meetings and better presentation layouts, as well as support for more devices in the future.

For the most part, having this many people stare at me from my screen doesn’t seem necessary (and more likely to induce stress than anything else), but the ability to present a single Chrome tab is surely a welcome new feature for many. But what’s probably just as important is that this means you can share higher-quality video content from these tabs than before.

If you often take meetings in the dark, low-light mode uses AI to brighten up your video. Unlike some of the other features, this one is coming to mobile first and will come to web users in the future.

Personally, I’m most excited about the new noise cancellation feature. Typically, noise cancellation works best for noises that repeat and are predictable. Think about the constant drone of an airplane or your neighbor’s old lawnmower. But Google says Meet can now go beyond this and also cancel out barking dogs and your noisy keystrokes. That has increasingly become table stakes, with even Discord offering similar capabilities and Nvidia RTX Voice now making this available in a slew of applications for users of its high-end graphics cards, but it’s nice to see this as a built-in feature for Meet now.

This feature will only roll out in the coming weeks and will initially be available to G Suite Enterprise and G Suite Enterprise for Education users on the web, with mobile support coming later.

AWS today launched Amazon AppFlow, a new integration service that makes it easier for developers to transfer data between AWS and SaaS applications like Google Analytics, Marketo, Salesforce, ServiceNow, Slack, Snowflake and Zendesk. Like similar services, including Microsoft Azure’s Power Automate, for example, developers can trigger these flows based on specific events, at pre-set times or on-demand.

Unlike some of its competitors, though, AWS is positioning this service more as a data transfer service than a way to automate workflows, and, while the data flow can be bi-directional, AWS’s announcement focuses mostly on moving data from SaaS applications to other AWS services for further analysis. For this, AppFlow also includes a number of tools for transforming the data as it moves through the service.

“Developers spend huge amounts of time writing custom integrations so they can pass data between SaaS applications and AWS services so that it can be analysed; these can be expensive and can often take months to complete,” said AWS principal advocate Martin Beeby in today’s announcement. “If data requirements change, then costly and complicated modifications have to be made to the integrations. Companies that don’t have the luxury of engineering resources might find themselves manually importing and exporting data from applications, which is time-consuming, risks data leakage, and has the potential to introduce human error.”

Every flow (which AWS defines as a call to a source application to transfer data to a destination) costs $0.001 per run, though, in typical AWS fashion, there’s also cost associated with data processing (starting at 0.02 per GB).

“Our customers tell us that they love having the ability to store, process, and analyze their data in AWS. They also use a variety of third-party SaaS applications, and they tell us that it can be difficult to manage the flow of data between AWS and these applications,” said Kurt Kufeld, vice president, AWS. “Amazon AppFlow provides an intuitive and easy way for customers to combine data from AWS and SaaS applications without moving it across the public internet. With Amazon AppFlow, our customers bring together and manage petabytes, even exabytes, of data spread across all of their applications — all without having to develop custom connectors or manage underlying API and network connectivity.”

At this point, the number of supported services remains comparatively low, with only 14 possible sources and four destinations (Amazon Redshift and S3, as well as Salesforce and Snowflake). Sometimes, depending on the source you select, the only possible destination is Amazon’s S3 storage service.

Over time, the number of integrations will surely increase, but for now, it feels like there’s still quite a bit more work to do for the AppFlow team to expand the list of supported services.

AWS has long left this market to competitors, even though it has tools like AWS Step Functions for building serverless workflows across AWS services and EventBridge for connections applications. Interestingly, EventBridge currently supports a far wider range of third-party sources, but as the name implies, its focus is more on triggering events in AWS than moving data between applications.

The pandemic may feel all-encompassing at the moment, but Confluent announced a $250 million Series E today, showing that major investment continues in spite of the dire economic situation at the moment. The company is now valued at $4.5 billion.

Today’s round follows last year’s $125 million Series D. At that point the company was valued at a mere $2.5 billion. Investors obviously see a lot of potential here.

Coatue Management led the round, with help from Altimeter Capital and Franklin Templeton. Existing investors Index Ventures and Sequoia Capital also participated. Today’s investment brings the total raised to $456 million.

The company is based on Apache Kafka, the open-source streaming data project that emerged from LinkedIn in 2011. Confluent launched in 2014 and has gained steam, funding and gaudy valuations along the way.

CEO and co-founder Jay Kreps reports that growth continued last year when sales grew 100% over the previous year. A big part of that is the cloud product the company launched in 2017. It added a free tier last September, which feels pretty prescient right about now.

But the company isn’t making money giving stuff away, so much as attracting users, who can become customers at some point as they make their way through the sales funnel. The beauty of the cloud product is that you can buy by the sip.

The company has big plans for the product this year. Although Kreps was loath to go into detail, he says that there will be a series of changes coming up this year that will add significantly to the product’s capabilities.

“As part of this we’re going to have a major new set of capabilities for our cloud service, and for open-source Kafka, and for our product that we’re going to announce every month for the rest of the year,” Kreps told TechCrunch. These will start rolling out the first week in May.

While he wouldn’t get specific, he says that it relates to the changing nature of cloud infrastructure deployment. “This whole infrastructure area is really evolving as it moves to the cloud. And so it has to become much, much more elastic and scalable as it really changes how it works. And we’re going to have announcements around what we think are the core capabilities of event streaming in the cloud,” he said.

While a round this big with a valuation this high and an institutional investor like Franklin Templeton involved typically means an IPO could be the next step, Kreps was not ready to talk about that, except to say the company does plan to begin behaving in the cadence of a public company with a set of quarterly earnings, just not for public consumption yet.

The company was founded in 2014. It has 1,000 employees and has plans to continue to hire and to expand the product. Kreps sees plenty of opportunity here in spite of the current economics.

“I don’t think you want to just turtle up and hang on to your existing customers and not expand if you’re in a market that’s really growing. What really got this round of investors excited is the fact that we’re onto something that has a huge market, and we want to continue to advance, even in these really weird uncertain times,” he said.

AWS and Facebook today announced two new open-source projects around PyTorch, the popular open-source machine learning framework. The first of these is TorchServe, a model-serving framework for PyTorch that will make it easier for developers to put their models into production. The other is TorchElastic, a library that makes it easier for developers to build fault-tolerant training jobs on Kubernetes clusters, including AWS’s EC2 spot instances and Elastic Kubernetes Service.

In many ways, the two companies are taking what they have learned from running their own machine learning systems at scale and are putting this into the project. For AWS, that’s mostly SageMaker, the company’s machine learning platform, but as Bratin Saha, AWS VP and GM for Machine Learning Services, told me, the work on PyTorch was mostly motivated by requests from the community. And while there are obviously other model servers like TensorFlow Serving and the Multi Model Server available today, Saha argues that it would be hard to optimize those for PyTorch.

“If we tried to take some other model server, we would not be able to quote optimize it as much, as well as create it within the nuances of how PyTorch developers like to see this,” he said. AWS has lots of experience in running its own model servers for SageMaker that can handle multiple frameworks, but the community was asking for a model server that was tailored toward how they work. That also meant adapting the server’s API to what PyTorch developers expect from their framework of choice, for example.

As Saha told me, the server that AWS and Facebook are now launching as open source is similar to what AWS is using internally. “It’s quite close,” he said. “We actually started with what we had internally for one of our model servers and then put it out to the community, worked closely with Facebook, to iterate and get feedback — and then modified it so it’s quite close.”

Bill Jia, Facebook’s VP of AI Infrastructure, also told me, he’s very happy about how his team and the community has pushed PyTorch forward in recent years. “If you look at the entire industry community — a large number of researchers and enterprise users are using AWS,” he said. “And then we figured out if we can collaborate with AWS and push PyTorch together, then Facebook and AWS can get a lot of benefits, but more so, all the users can get a lot of benefits from PyTorch. That’s our reason for why we wanted to collaborate with AWS.”

As for TorchElastic, the focus here is on allowing developers to create training systems that can work on large distributed Kubernetes clusters where you might want to use cheaper spot instances. Those are preemptible, though, so your system has to be able to handle that, while traditionally, machine learning training frameworks often expect a system where the number of instances stays the same throughout the process. That, too, is something AWS originally built for SageMaker. There, it’s fully managed by AWS, though, so developers never have to think about it. For developers who want more control over their dynamic training systems or to stay very close to the metal, TorchElastic now allows them to recreate this experience on their own Kubernetes clusters.

AWS has a bit of a reputation when it comes to open source and its engagement with the open-source community. In this case, though, it’s nice to see AWS lead the way to bring some of its own work on building model servers, for example, to the PyTorch community. In the machine learning ecosystem, that’s very much expected, and Saha stressed that AWS has long engaged with the community as one of the main contributors to MXNet and through its contributions to projects like Jupyter, TensorFlow and libraries like NumPy.