Adobe is looking to get schooled on AI and data science.
While many technology giants foster relationships with academics by offering them lucrative part-time consultancy positions. Adobe is pursuing a different tack: dishing out $50,000 no-strings-attached grants to professors and doctoral students working on projects of joint interest.
“What academia provides is more the advanced mathematical algorithms and the advanced research that’s gone into other related areas but hasn’t been applied to our field,” said Anil Kamath, Adobe’s VP of technology.
Artificial Intelligentsia
Adobe was not an early promoter of AI products as were other major technology players, like Google with its Automated Insights pattern-recognition tool, IBM with Watson and Einstein from Salesforce.
But Adobe’s research grant program, which has dished out 40 grants for a total of $2 million in the past four years, is bringing algorithmic AI into the company through academic work.
Adobe is also doing outreach at events. A top priority at its fourth annual Data Science Symposium held in San Jose last week was to identify AI and machine learning research proposals.
“We want to take the areas of focus for data scientists working on AI and machine learning and funnel them to real-world digital marketing problems,” Kamath said.
For instance, Rutgers University computer science professor Shan Muthukrishnan, developed an algorithm that takes the “hundreds of dimensions” of audience data coming into a cloud – cookies, browser data, device IDs, audience, demographic data points, location, et al. – and learns to pluck potentially actionable marketing trends from the raw data stream.
Adobe doesn’t own Muthukrishnan’s research, the grant is considered a hands-off gift to the university and the work that comes out of the research process is open to public and peer review.
But Adobe does provide the data and in that is able to point the professor’s research at thorny product issues facing the company.
Productizing Research
What is Adobe getting out of this? It doesn’t own the research or the algorithms being developed and researchers from Oracle or Salesforce could likewise read the theses and academic journal papers that result from Adobe research grants.
One way Adobe capitalizes on its grants is by connecting major customers with academic researchers to help them solve their big data problems.
Alice Li worked with Adobe as part of her doctoral research in 2012 and more recently won a grant from Adobe to work on an attribution model for an online jewelry retailer – and Adobe client – meant to bridge the gap between last-touch and multitouch attribution.
The jeweler ended up deploying the attribution model, and continues to use it as part of its work with Adobe.
“The company is judging the marginal value of their marketing channels based on what they’re told by those marketing channels,” Li said. “Google tells them paid search is very effective, Facebook tells them social is very effective, but academic research will be objective and rigorous.”
And although the grants don’t give Adobe intellectual property rights or new software, they do get productized through human capital, namely interns and cross-employed researchers.
This past year Adobe awarded a grant to a Stanford professor and graduate student working on sequential recommendations, specifically: How should a platform lead users through video tutorial sequences based on factors like whether that person is on a free version of a product or already a paid subscriber? And how likely is the user to churn and abandon the platform entirely?
The graduate student is employed over the summer by Adobe, where – surprise, surprise – he’ll be working on its video tutorial sequencing.
“Usually the PhD students working in that area have a chance to work with us,” Kamath said. “And that’s where we see some of the best results translating research work and putting it into product.”
Talent Funnel
But even more important that capitalizing on research to inform product development is the chance to secure long-term talent.
“A part of it is getting academia and researchers to think about and work on problems that are relevant to us,” Kamath said. “[But] another big part is recruiting these machine learning and data science students who are really competitively sought after.”
It’s good for students to consider industry concerns, said Stanford professor Ramesh Johari, who’s worked on multiple Adobe research grants, including the sequential learning algorithm.
“Students can benefit by being aware of the connections between the algorithmic work they’re doing and actual problems people face,” he said.
He’s seen students receive their masters or PhD and go on to corporate tech development in order to develop the work they did in school. Adobe researchers have also served as co-authors on academic projects and sat on thesis defense committees.
The advantages of academic outreach aren’t immediate and are hard to quantify with a dollar sign, aside from the batch of $50,000 checks the company awards each year, of course. But the long-term advantages are clear.
“Adobe is really well served by work intersecting with the academic community,” Johari said.
