data science
Auto Added by WPeMatico
DataRobot is acquiring Paxata to add data prep to machine learning platform
DataRobot, a company best known for creating automated machine learning models known as AutoML, announced today that it intends to acquire Paxata, a data prep platform startup. The companies did not reveal the purchase price.
Paxata raised a total of $90 million before today’s acquisition, according to the company.
Up until now, DataRobot has concentrated mostly on the machine learning and data science aspect of the workflow — building and testing the model, then putting it into production. The data prep was left to other vendors like Paxata, but DataRobot, which raised $206 million in September, saw an opportunity to fill in a gap in their platform with Paxata.
“We’ve identified, because we’ve been focused on machine learning for so long, a number of key data prep capabilities that are required for machine learning to be successful. And so we see an opportunity to really build out a unique and compelling data prep for machine learning offering that’s powered by the Paxata product, but takes the knowledge and understanding and the integration with the machine learning platform from DataRobot,” Phil Gurbacki, SVP of product development and customer experience at DataRobot, told TechCrunch.
Prakash Nanduri, CEO and co-founder at Paxata, says the two companies were a great fit and it made a lot of sense to come together. “DataRobot has got a significant number of customers, and every one of their customers have a data and information management problem. For us, the deal allows us to rapidly increase the number of customers that are able to go from data to value. By coming together, the value to the customer is increased at an exponential level,” he explained.
DataRobot is based in Boston, while Paxata is in Redwood City, Calif. The plan moving forward is to make Paxata a west coast office, and all of the company’s almost 100 employees will become part of DataRobot when the deal closes.
While the two companies are working together to integrate Paxata more fully into the DataRobot platform, the companies also plan to let Paxata continue to exist as a standalone product.
DataRobot has raised more than $431 million, according to PitchBook data. It raised $206 million of that in its last round. At the time, the company indicated it would be looking for acquisition opportunities when it made sense.
This match-up seems particularly good, given how well the two companies’ capabilities complement one another, and how much customer overlap they have. The deal is expected to close before the end of the year.
Powered by WPeMatico
Neural Magic gets $15M seed to run machine learning models on commodity CPUs
Neural Magic, a startup founded by a couple of MIT professors, who figured out a way to run machine learning models on commodity CPUs, announced a $15 million seed investment today.
Comcast Ventures led the round, with participation from NEA, Andreessen Horowitz, Pillar VC and Amdocs. The company had previously received a $5 million pre-seed, making the total raised so far $20 million.
The company also announced early access to its first product, an inference engine that data scientists can run on computers running CPUs, rather than specialized chips like GPUs or TPUs. That means that it could greatly reduce the cost associated with machine learning projects by allowing data scientists to use commodity hardware.
The idea for this solution came from work by MIT professor Nir Shavit and his research partner and co-founder Alex Mateev. As he tells it, they were working on neurobiology data in their lab and found a way to use the commodity hardware he had in place. “I discovered that with the right algorithms we could run these machine learning algorithms on commodity hardware, and that’s where the company started,” Shavit told TechCrunch.
He says there is this false notion that you need these specialized chips or hardware accelerators to have the necessary resources to run these jobs, but he says it doesn’t have to be that way. He says his company not only allows you to use this commodity hardware, it also works with more modern development approaches, like containers and microservices.
“Our vision is to enable data science teams to take advantage of the ubiquitous computing platforms they already own to run deep learning models at GPU speeds — in a flexible and containerized way that only commodity CPUs can deliver,” Shavit explained.
He says this also eliminates the memory limitations of these other approaches because CPUs have access to much greater amounts of memory, and this is a key advantage of his company’s approach over and above the cost savings.
“Yes, running on a commodity processor you get the cost savings of running on a CPU, but more importantly, it eliminates all of these huge commercialization problems and essentially this big limitation of the whole field of machine learning of having to work on small models and small data sets because the accelerators are kind of limited. This is the big unlock of Neural Magic,” he said.
Gil Beyda, managing director at lead investor Comcast Ventures, sees a huge market opportunity with an approach that lets people use commodity hardware. “Neural Magic is well down the path of using software to replace high-cost, specialized AI hardware. Software wins because it unlocks the true potential of deep learning to build novel applications and address some of the industry’s biggest challenges,” he said in a statement.
Powered by WPeMatico
Vianai emerges with $50M seed and a mission to simplify machine learning tech
You don’t see a startup get a $50 million seed round all that often, but such was the case with Vianai, an early-stage startup launched by Vishal Sikka, former Infosys managing director and SAP executive. The company launched recently with a big check and a vision to transform machine learning.
Just this week, the startup had a coming out party at Oracle Open World, where Sikka delivered one of the keynotes and demoed the product for attendees. Over the last couple of years, since he left Infosys, Sikka has been thinking about the impact of AI and machine learning on society and the way it is being delivered today. He didn’t much like what he saw.
It’s worth noting that Sikka got his PhD from Stanford with a specialty in AI in 1996, so this isn’t something that’s new to him. What’s changed, as he points out, is the growing compute power and increasing amounts of data, all fueling the current AI push inside business. What he saw when he began exploring how companies are implementing AI and machine learning today was a lot of complex tooling, which, in his view, was far more complex than it needed to be.
He saw dense Jupyter notebooks filled with code. He said that if you looked at a typical machine learning model, and stripped away all of the code, what you found was a series of mathematical expressions underlying the model. He had a vision of making that model-building more about the math, while building a highly visual data science platform from the ground up.
The company has been iterating on a solution over the last year with two core principles in mind: explorability and explainability, which involves interacting with the data and presenting it in a way that helps the user attain their goal faster than the current crop of model-building tools.
“It is about making the system reactive to what the user is doing, making it completely explorable, while making it possible for the developer to experiment with what’s happening in a way that is incredibly easy. To make it explainable means being able to go back and forth with the data and the model, using the model to understand the phenomenon that you’re trying to capture in the data,” Sikka told TechCrunch.
He says the tool isn’t just aimed at data scientists, it’s about business users and the data scientists sitting down together and iterating together to get the answers they are seeking, whether it’s finding a way to reduce user churn or discover fraud. These models do not live in a data science vacuum. They all have a business purpose, and he believes the only way to be successful with AI in the enterprise is to have both business users and data scientists sitting together at the same table working with the software to solve a specific problem, while taking advantage of one another’s expertise.
For Sikka, this means refining the actual problem you are trying to solve. “AI is about problem solving, but before you do the problem solving, there is also a [challenge around] finding and articulating a business problem that is relevant to businesses and that has a value to the organization,” he said.
He is very clear, that he isn’t looking to replace humans, but instead wants to use AI to augment human intelligence to solve actual human problems. He points out that this product is not automated machine learning (AutoML), which he considers a deeply flawed idea. “We are not here to automate the jobs of data science practitioners. We are here to augment them,” he said.
As for that massive seed round, Sikka knew it would take a big investment to build a vision like this, and with his reputation and connections, he felt it would be better to get one big investment up front, and he could concentrate on building the product and the company. He says that he was fortunate enough to have investors who believe in the vision, even though as he says, no early business plan survives the test of reality. He didn’t name specific investors, only referring to friends and wealthy and famous people and institutions. A company spokesperson reiterated they were not revealing a list of investors at this time.
For now, the company has a new product and plenty of money in the bank to get to profitability, which he states is his ultimate goal. Sikka could have taken a job running a large organization, but like many startup founders, he saw a problem, and he had an idea how to solve it. That was a challenge he couldn’t resist pursuing.
Powered by WPeMatico
Y Combinator-backed Narrator wants to become the operating system for data science
Cedric Dussud, Michael Nason, Ahmed Elsamadisi and Matthew Star (pictured above, in order) spent the summer sharing a house in San Francisco, cooking meals together and building Narrator, a startup with ambitions of becoming a universal data model fit for any company.
Narrator is one of more than 100 startups graduating next week from Y Combinator, the San Francisco accelerator program. Put simply, the company provides data-science-as-a-service to its customers: fellow startups.
“We provide the equivalent of a data team for the price of an analyst,” explains Narrator co-founder and director of engineering Star. “Within the first month, our clients get an infinitely scalable data system.”
Led by chief executive officer Elsamadisi, a former senior data engineer at WeWork, the Narrator founding team is made up entirely of alums of the co-working giant. The building blocks of Narrator’s subscription-based data modeling tool were developed during Elsamadisi’s WeWork tenure, where he was tasked with making sense of the company’s disorganized trove of data.
As an early addition to WeWork’s data team, Elsamadisi spent two years bringing WeWork’s data to one place, scaling the team to 40 people and ultimately creating a functional data model the soon-to-be-public company could use to streamline operations. Then in 2017, Elsamadisi had an a-ha moment. The system he created at WeWork could be applied to any data stream, he thought.
“All companies are fundamentally the same when it comes to the kinds of data they want to understand about their business,” Narrator’s Dussud tells TechCrunch. “Every startup wants to know what’s my monthly recurring revenue, why are my customers churning or whatever the case may be. The only reason they have to go hire a data team and hire a business analyst is because the way that their data is structured is specific to that company.”
All Narrator clients use the same consistent format to absorb and manage their data, saving startups time and heaps of money.
Narrator follows a long line of Y Combinator graduates that built startups catering to other startups, as the accelerator becomes more of a SaaS incubator of sorts. PagerDuty and Docker proved that YC companies could build with a strong focus on other YC companies. Brex, a recent YC grad that issues credit cards to entrepreneurs, has leveraged the same startup-focused model for big-time success.
“Why not build a company to make something that other startups can have?” Asks Dussud. “It’s hugely valuable and only big companies have access to it. Let’s make it available to everybody.”
New York-based Narrator sees a massive opportunity ahead. Every company, after all, wants to increase revenue or decrease costs, a difficult task easier accomplished with a data-driven culture.
“If you start to imagine a world where, under the hood, the structure of the data at all companies is the same, you can now start reusing a lot of the things that in the past would actually be quite complicated,” said Star. “Right now, anytime you want to start from scratch with a new data system, you are literally starting from scratch and unfortunately reinventing the wheel. If you had a standardized system, you know, a standardized model, you could start reusing a lot of really wonderful things.”
Narrator is working with 14 clients today, each using an identical data model. Their goal is for Narrator’s structure to become the standard by which all startups do data science. In other words, Narrator hopes to become the operating system for data science.
“What’s kind of amazing is whether we’re working with a financial app … a clothing rental startup or a healthcare company, they’re all using the same data model,” said Star. “Any one of those teams, if they wanted to get the same level of analysis, they would have to hire a data analyst.”
Narrator raised $1.3 million in seed funding led by Flybridge Capital Partners prior to joining YC. Hot off the heels of the accelerator program, there’s no doubt the startup will close another round of financing soon.
Powered by WPeMatico
Adobe’s latest Customer Experience Platform updates take aim at data scientists
Adobe’s Customer Experience Platform provides a place to process all of the data that will eventually drive customer experience applications in the Adobe Experience Cloud. This involves bringing in vast amounts of transactional and interactional data being created across commerce platforms. This process is complex and involves IT, applications developers and data scientists.
Last fall, the company introduced a couple of tools in beta for the last group. Data scientists need familiar kinds of tools to work with the data as it streams into the platform in order to create meaningful models for the application developers to build upon. Today, it made two of those tools generally available — Query Service and Data Science Workspaces — which should go a long way toward helping data scientists feel comfortable working with data on this platform.
Ronell Hugh, group manager at Adobe Experience Platform, says these tools are about helping data scientists move beyond pure data management and getting into deriving more meaningful insights from it. “Data scientists were just bringing data in and trying to manage and organize it, and now we see that with Experience Platform, they are able to do that in a more seamless way, and can spend more time doing what they really want to do, which is deriving insights from the data to be actionable in the organization,” Hugh told TechCrunch.
Part of that is being able to do queries across the data sets they have brought into the platform. The newly released Query Service will enable data scientists and analysts to write queries to understand the data better and get specific answers based on the data faster.
“With Query Service in Adobe Experience Platform, analysts and data scientists can now poll all of their data sets stored in Experience Platform to answer specific cross-channel and cross-platform questions, faster than ever before. This includes behavioral data, as well as point-of-sale (POS), customer relationship management (CRM) and more,” the company wrote in a blog post announcing the new tool.
In addition, the company made the Data Science Workspace generally available. As the name implies, it provides a place for data scientists to work with the data and build models derived from it. The idea behind this tool is to use artificial intelligence to help automate some of the more mundane aspects of the data science job.
“Data scientists can take advantage of this new AI that fuels deeper data discovery by using Adobe Sensei pre-built models, bringing their existing models or creating custom models from scratch in Experience Platform,” the company wrote in the announcement blog post.
Today, it was the data scientists’ turn, but the platform is designed to help IT manage underlying infrastructure, whether in the cloud or on premises, and for application developers to take advantage of the data models and build customer experience applications on top of that. It’s a complex, yet symbiotic relationship, and Adobe is attempting to pull all of it together in a single platform.
Powered by WPeMatico
What we can learn from DTC success with TV ads
Kevin Krim is EDO‘s President & CEO. His 21-year career has spanned search, social and TV advertising across start-ups and major companies like Yahoo and NBCUniversal. Sebastian Chiu is EDO‘s Chief Data Scientist. He earned his undergraduate and post-graduate degrees from Harvard, working previously as a data scientist at Dropbox.
One of the most-discussed plot twists in recent advertising has been the pivot of Direct-to-Consumer (DTC) brands to linear TV. These data-driven, digital-first players are expanding well beyond Facebook and Instagram—and becoming serious players on the largest traditional medium in advertising.
A January 2019 Video Advertising Bureau study found that in 2018, 120 DTC brands collectively spent over $2 billion in TV ads—up from $1.1 B in 2016. 70 of those 2018 advertisers ran TV ads for the first time.
But while we know that they’re advertising on TV, what may be less discussed is whether they’re succeeding on television—and what strategies they use to achieve their success.
At EDO, we have a unique and differentiated ability to measure how DTC advertisers perform on TV by tracking incremental online searches above baseline in the minutes immediately following individual TV ad airings as viewers translate their interest in advertised brands and products directly into online engagement with them.
By measuring incremental search activity across 60 million national TV ad airings since 2015, we are able to effectively isolate the effects of TV ad placement and creative decisions that are most likely to cause online engagement.
We ran the numbers on DTCs as well as advertisers in various other categories to better understand how DTCs specifically are succeeding in TV ads—and what DTCs who are considering TV advertising can do to achieve success on TV.
Table of Contents
- Does the David vs Goliath story play out on TV?
- Is there a newcomer advantage on TV?
- How does the DTC marketing playbook translate to TV? (A brief history and how-to, plus a Casper mini-case study)
- Conclusion: 3 keys to DTC TV success
Does the David vs. Goliath story play out on TV?
The DTC revolution is a quintessential David and Goliath story. In vertical after vertical, small, digital-native upstarts are changing the game and overtaking major brands. Does that story play out on TV as well—or is TV advertising one area where DTC marketers have finally met their match?
To answer that question, EDO looked at how effectively TV ads elicited viewer activity since September 2018 across eight major industry categories including DTC. Guided by historical ad performance across billions of ads, we rated ad performance based on how closely the DTC ads came to meeting the benchmark volume of brand-related online activity in the minutes following each TV ad airing.
We index each industry accordingly—giving an index value of 100 to an ad that meets benchmark standards, and below-par ads getting a score under 100 while higher-scoring ads receive a score over 100. We chose to set our index baseline of 100 to the average Consumer Packaged Good (CPG) ad since it is such a large and broad ad category. Our results are as follows:
Powered by WPeMatico
Sisense acquires Periscope Data to build integrated data science and analytics solution
Sisense announced today that it has acquired Periscope Data to create what it is calling a complete data science and analytics platform for customers. The companies did not disclose the purchase price.
The two companies’ CEOs met about 18 months ago at a conference, and running similar kinds of companies, hit it off. They began talking and, after a time, realized it might make sense to combine the two startups because each one was attacking the data problem from a different angle.
Sisense, which has raised $174 million, tends to serve business intelligence requirements either for internal use or externally with customers. Periscope, which has raised more than $34 million, looks at the data science end of the business.
Both CEOs say they could have eventually built these capabilities into their respective platforms, but after meeting they decided to bring the two companies together instead, and they made a deal.
Harry Glaser from Periscope Data and Amir Orad of Sisense
“I realized over the last 18 months [as we spoke] that we’re actually building leadership positions into two unique areas of the market that will slowly become one as industries and technologies evolve,” Sisense CEO Amir Orad told TechCrunch.
Periscope CEO Harry Glaser says that as his company built a company around advanced analytics and predictive modeling, he saw a growing opportunity around operationalizing these insights across an organization, something he could do much more quickly in combination with Sisense.
“[We have been] pulled into this broader business intelligence conversation, and it has put us in a place where as we do this merger, we are able to instantly leapfrog the three years it would have taken us to deliver that to our customers, and deliver operationalized insights on integration day on day one,” Glaser explained.
The two executives say this is part of a larger trend about companies becoming more data-driven, a phrase that seems trite by now, but as a recent Harvard Business School study found, it’s still a big challenge for companies to achieve.
Orad says that you can debate the pace of change, but that overall, companies are going to operate better when they use data to drive decisions. “I think it’s an interesting intellectual debate, but the direction is one direction. People who deploy this technology will provide better care, better service, hire better, promote employees and grow them better, have better marketing, better sales and be more cost effective,” he said.
Orad and Glaser recognize that many acquisitions don’t succeed, but they believe they are bringing together two like-minded companies that will have a combined ARR of $100 million and 700 employees.
“That’s the icing on the cake, knowing that the cultures are so compatible, knowing that they work so well together, but it starts from a conviction that this advanced analytics can be operationalized throughout enterprises and [with] their customers. This is going to drive transformation inside our customers that’s really great for them and turns them into data-driven companies,” Glaser said.
Powered by WPeMatico
Gamalon scores $20 M led by Intel Capital
Gamalon wants to change the game when it comes to understanding text-based customer communications. Instead of using neural networks to learn about a vast corpus of information, the startup takes a different approach, putting the text in a database and building decision trees to very rapidly train the data to arrive at the required information. Today, it announced a $20 million Series A investment led by Intel Capital.
Other participants in the round included .406 Ventures and Omidyar Technology Ventures along with existing investors Boston Seed Capital, Felicis Ventures and Rivas Capital. Today’s investment brings the total raised by the company since inception in 2013 to $32 million including backing from DARPA in earlier rounds.
Gamalon CEO Ben Vigoda says they developed a new approach to analyzing customer interactions because the state of the art in AI and machine learning was too much of a black box.
His company wants to change that by making the whole process much more interactive. To that end Gamalon also released a new tool called Idea Studio, a product that can automatically build learning trees to help users arrive at answers extremely fast or allow a business analyst or data scientists to simply enter a series of queries and build a decision tree on the fly based on the text. With neural networks, Vigoda says, the user has no control over the end result, but with Idea Studio you can edit the trees and refine the results immediately.
Gamalon Idea Studio decision tree. Photo: Gamalon
The product still needs a way to review all of the text-based content, of course, but instead of having humans categorize it all manually, with Gamalon you import your data into a database, do analytics on it and then make it available for rapid categorization and response.
This could have multiple utilities, whether for customer service agents to find answers very quickly or customers to interact with bots and find answers much faster. Analysts could use it to locate answers to business issues, and it’s sophisticated enough for data scientists to build machine learning projects based on a large corpus of data.
You can build a learning tree by entering related text to train it. GIF: Gamalon
Naveen Rao, corporate vice president and general manager in the Artificial Intelligence Products Group at Intel Corporation says they like how Gamalon puts machine learning into hands of many different employees around the customer information use case. “We want enterprises of all levels of AI capability to take full advantage of this growing volume and complexity of data. Gamalon’s unique approach can help users better understand billions of customer communications, customize individual responses, and take action to better serve those customers,” Rao explained in a statement.
The company is based in Cambridge, MA and has 23 employees. They have six large customers including Avaya.
Powered by WPeMatico
Feedzai closes $50M Series C to help banks and merchants identify fraud with AI
Feedzai is announcing a $50 million Series C this morning led by an unnamed VC with additional capital from Sapphire Ventures. The six year old startup builds machine learning tools to help banks and merchants spot payment fraud. In today’s rapidly maturing world of fintech, Feedzai is trying to thread the needle between turnkey solution and customizable platform. With 60 clients… Read More
Powered by WPeMatico
Alteryx Promote puts data science to work across the company
When Alteryx acquired Yhat in June, it was only a matter of time before the startup’s data-science management software began showing up in Alteryx. Just today, the company announced Alteryx Promote, a new tool based on Yhat’s product set.
The company made the announcement at the Alteryx Inspire Europe customer event taking place in London this week.
Alteryx Promote gives data… Read More
Powered by WPeMatico