Unifying Data Across Silos: Tamr’s Anthony Deighton

Welcome to the software report, a leading information source on software companies and solutions. In this episode, we speak with Anthony Data, CEO of Tamer, an AI native master data management solution that unifies, cleans, and enriches fragmented enterprise data to produce golden records for generative AI initiatives, automation, and decision-making. Tamer's patented approach combines machine learning and AI agents with human refinement and oversight, delivering value in days or weeks and giving organizations 360 degree views that link data across silos in real time. Anthony previously served as Tamer's chief product officer and general manager for data products, and earlier held senior leadership roles at Salonis and click focused on enterprise data and software growth. I'm your host, RJ Lomba. We hope you enjoy the show. If you like the episode, click to follow. Anthony, thank you so much for taking the time. Delight to be with you. Oh, it's a total pleasure, thanks. Where I think we could kick off is for the benefit of our audience. You operate in the data space. Obviously, that's a very important space in the age of AI. Let's start off with a quick kind of overview of Tamer. Sure. Tamer is in the master data management space, and you're right about your point about AI. What's interesting from my perspective around Tamer and master data management is that as someone so eloquently put it to me recently, MDM is having a moment. I think that's a function of organizations increasingly deploying these AI workloads, whether they're AI agents or AI models. And they're realizing that their data is a mess, right? It's unorganized. It's in silos. It's messy. It's incomplete. It's incorrect. And then when you turn that over to an AI agent, bad things can happen. Bad things can happen at scale, and that's scary. And so companies are saying, well, geez, we've got to get a handle on this. We've got to organize this data, collect it together, and we want to do that across systems. We want to be thinking across business units, across geographies, across product lines, we need to organize the data in a different way. MDM can help with that. And then Tamer can be a part of that solution. This also speaks to an important idea, which is Tamer uses AI to solve the MDM problems. So we're both being driven by this drive to AI inside the organization and the beneficiary of it. And there's others in the MDM space. You have a unique history, and then I believe you grew out of MIT, and you have a notable founder, maybe give us a little bit of that history. Sure. So Tamer came out of academic research at MIT in the computer science department under the tutelage of Mike Stonebreaker, and Stonebreaker is obviously turning award-winner a very famous guy. And he had a team of folks in his lab, and they were working on it in a way, what sounds like a very simple problem, which is, can we train a model, we train a machine to do basic data management tasks? So imagine you have data in lots of places, we want to link it together, we might consider doing that manually, or we might consider doing that with rules through a rules engine. What if we could train a model to do the same thing? And it started really with just that computer science problem, that generated an academic paper, and then from that academic paper, the original idea behind the tool was data Tamer TAM ER, and then of course somebody realized there was a copyright violation, so it became Tamer TAM. But again, core idea that came out of that academic research is still very relevant today. What if we could have a machine do the work we would traditionally rely on humans to do around organizing, cleaning, updating, keeping up to date, enterprise data. That's really the core idea. And fast forward to today, what types of companies use your solution, are these larger companies that have a lot of data that they need organized? Before we talk about the customer, you said fast forward to today, and this is actually a really important idea, which is when the original academic work was done, this idea that we could rely on machines to do work that we would traditionally rely on humans to do, was kind of fanciful for being honest. And I call it the chat GPT moment, but as soon as AI hit the mainstream, as soon as my mother-in-law started using it, all of a sudden this idea that we could rely on a machine to do this task, or these tasks, all of a sudden became like, yeah, this is actually possible. And I think that's a lot of what drove the interest of adoption of Tamer's approach. So again, using the machine to organize the data and not rules or people. But to your question, yes, generally we see the larger the company, like more data, more problems, like the bigger the company, the more likely they are to have grown through acquisition to operate in geographic divisions or product lines to have many different underlying systems to be organized with different data warehousing technologies, and then struggle with this challenge of knitting that data together. What I've found, and this is somewhat anecdotal though, is that there's still a little bit of error. If I use chat GPT, I have a degree of confidence that what it's reporting back to me is accurate, right? But when you're applying AI at such a large scale. And for such an important thing, which is your enterprise data, for sure. How do you kind of manage that kind of confidence level? Like, and do most companies who use Tamer say, yeah, this is 99.9% accurate? So the first thing to remember is that let's assume for the moment that just manually going through the data is your alternative approach, right? That also has an error rate, right? People are fallible. Or if you imagine trying to write rules to do the same, again, rules aren't perfect either. And there's a reason people turn away from a rules-based approach. They're hard to manage. Rules can be conflicting. They can result in outcomes you didn't expect because two rules become in conflict with each other. So there is no perfection. That being said, one of the powerful ideas behind a model is that models by their nature are probabilistic. And so it's much easier for a system like Tamer that's built with models at the core to surface the cases in which the model is least confident. So with a rule, there's no sense of confidence. It's like it fired or it didn't, right? With Tamer, we can say these are cases where the model's having the most trouble. And these are likely places where you as the human want to make a tough call and make a decision. We have this concept we call the curator hub. And you can think of it almost like as a cue of places where the human can be most helpful. And this also speaks to the manual approach. If I asked you to go through 10 million records manually, you would probably quit. You'd be like, "I'm not doing that. I'm life's too short." But if I said, "Look, here's 10 records that I'm having the most trouble with that would have the biggest impact on improving your business." You'd be like, "I'll do that. 10 records got it. No problem." And that's the distinction. We can apply human energy in the place where it's most valuable. Let's talk about the application of the solutions. I believe healthcare is one of the big sectors. So maybe a great case where Tamer was put to use. Yeah, so I think about an organization like CHG, which is in the healthcare staffing space. And if you think about it, this is a kind of a perfect problem. If you're trying to match clinicians to open roles, you often have many different systems that are contributing information to that match. You might have hospital systems with healthcare data. You also have physician systems. You may have background. You may have government data like NPI numbers. Not to mention the fact that you have an open role with a specific set of requirements. That's an whole different system. You might have other operational systems that are managing bits and pieces of the enrollment. And more importantly, matching the physician to the open role is really important. It's important to the physician, obviously. And it's important to the place they're going to go be working and be helpful. You want to make sure that they have admitting privileges that they're the right skill set, et cetera. And that's exactly the challenge that we're helping CHG healthcare solve. It's also not a small problem. There are millions of physicians running through their data sets and their network, literally over 7 million. And there are many different systems involved. And so you could end up with lots of duplicate data. But putting Tamer in place, CHG now has this 360 view of the provider in this case and thinking about what's the perfect place, perfect role for that provider at this moment. Maybe switching gears a little bit back to Tamer as a company. What stage are you at in terms of the life cycle? We talked about the origins and then we fast forward it all the way today. But can you give us a sense of the scale of the business and where you view it's at? Yeah, so I think we're in a way early in our journey. The original academic work was very much just that academic work. There's a big difference between a set of algorithms and academic paper and then building a company. Now, fast forward, the good news is we're 60 people, almost all of whom are here in Harvard Square, Cambridge, Massachusetts, we have a team in London. as well, I'm going to few folks out in Asia Pacific, but the bulk of people are here. And really the focus of the energy of the business is around making the algorithms and ideas and that really the technology is super easy to use. And again, I think to borrow from ChatGPT for a second, I think this is the magic of ChatGPT and similar technologies, which is it took something that was available for actually for many years prior to that point and just made it really easy to access. And in that same way, we want to make the powerful AI that we've built for doing common data management tasks so easy. Anyone can just, you know, sort of sign up, provision, and run some data through the system, have it available, have it available in real time, have it available in their common analytical systems, make it like real easy. And you brought in capital along the way, you have some investors in terms of capital, are you in the burning phase, are you looking for additional growth capital at a certain point in time? Sure. So any A and Google Ventures are the primary backers of the business. You know, we're at a place today where I think we control our destiny from a growth perspective. We don't need additional capital. Of course, I do think AI applications, if MDM is having a moment, AI applications are probably also having a moment. And I do think we're at an interesting inflection point in the business where we've reached this point where people trust the technology. They believe and understand that models can be helpful to them. And the central challenge, the thing we've been working on really for a long time around organizing enterprise data into these 360 pages that everyone can access, like that's kind of of the moment. So, you know, we're always looking for ways to fix all right, growth. I want to touch on your background. You have an interesting background in that you've served in different roles with different companies previously. And then when you entered tamer, it was as a chief product officer, I believe. So I'd love to hear a little bit about your background. I believe it begins at AT Karni. That's true. I began in management consulting for which I apologize. You know, I was a computer geek at heart and really wanted to work in the tech industry. I worked with Tom Siebel, Pat House, Dave Schmeier, the team at Siebel Systems when they were building that business. And I think it's fair to say that, you know, I learned an enormous amount from working at Siebel about what it means to be working at a professional company developing mission, critical, really important software for customers. Love that. Learned a lot from it. Then I joined a small, unknown, Swedish software company called Click. Literally, I was one of the first US employees. We helped build that business globally. We took a public. I started by running marketing and then ended up running product and actually took over engineering for a little while. And this is a common theme that I've had is this oscillation between the marketing and positioning side of the business and the product and engineering side of the business. And my general experience has been that it's really important to have both views. What is the market care about? What are customers talking about? How do they think about and use technology? And then what are we actually building? What is the core essence of differentiation that we can bring to the market? And I joined Salonus again, went back to the marketing side, running marketing there. It's a common theme between Salonus and Click was helping people visualizing consume data in new ways, whether it's Salonus or the process, visualization, or Click with a dashboard. But the common experience across both was the customers would complain not about the visualization but about the data itself. And they were like, I love this, but the data is junk. How do I fix that? And I've known Tamer since it was founded, but I sort of bumped into it. And I thought, you know what? The next 10 years in the data space are not going to be about helping people see and use data differently. It's going to be about actually improving the data. Like we've solved that visualization problem. Let's work on the data problem. And so that's why I joined here. I joined as cheap product officer because we were some product work we wanted to do. And then I've taken over and just running the business. So that's been fun too. This is an aside, but a lot of the younger generation now, computer engineers are having a tough time getting positions because they're being displaced by AI. So the question is, what fields should they go into? And data seems like it's a good field. Do you agree with that? I think that's right. And to say it a different way, the basic mechanics of writing code are becoming increasingly easy to turn over to machines. But the question and challenge associated with what do we want that code to do? What problem are we trying to solve? Is not one that I expect machines are going to be that helpful. That's the kind of experience that requires talking to a person, understanding their needs and requirements. It is also true that I think that hits first in the data space. So to say a different way, writing a Python pipeline to do a bunch of data transformation like cloud code can do that almost perfectly, understanding why you're transferring that data and what manipulation you're trying to do to it is a very different question. And when we're called, might have many suggestions, but wouldn't really be able to answer the question. Thinking about how you would express the results of that code you wrote, that's a very different problem. And to your point, my advice to someone starting their career would be to point yourself in the position of understanding a business problem, understanding a business person's challenges, and then translate that into code as opposed to starting with code and then try to find an application for. Right. There seems to be universities that are dedicating large amounts of capital to data sciences. So last two questions. One is, can you tell us about a person who has had a profound influence on you? Sure. I'm going to cheat and say two in probably the easy answer to this, which is my parents. But in a weird way, they both influence me in very similar ways. Both my parents are academics. My mother was a, where he has an English teacher, English professor. And so in a funny way, I think a lot about my marketing side comes from her thinking about the precision of language, the right sorts of words that meaning really matters as something I think she always was both very good at and sort of helped bring out in me. And then my dad is also an academic. And he, his retire, but he was a professor across the river at the Harvard Business School. And you know, one of the things about HBS is they use this case method and sort of the essence of the case method is whatever you say, I say the opposite, right? Like we create this sense of questioning everything. And one of the things I appreciate about my dad is an ability and willingness to just take a contrarian view just for the purpose of understanding the decision space. And that's always been something I've found to be very useful and learned from also maybe to say, both my parents are immigrants came to this country as did I suppose. So I will all of us have a lot of hope and respect for the kinds of opportunities that the United States sort of uniquely offers people who are here. So it's also true. Last question. Can you tell us about a charity cause or other endeavor that you're passionate about? Sure. This may be slightly unexpected. I for gosh, a long time now, over 10 years, I've participated in a local group called Newton Family Singers, which is an intergenerational chorus. I am not a good singer, which is to say, now after 10 years, I'm passable, I suppose. But it's certainly not something I would have looked at as a core strength of mine 10 years ago. But what I've appreciated about doing that is a couple of things. I'm not a naturally good singer. And yet this group of people can be accepting of somebody who's learning and trying something new. It's also intergenerational. My kids have done it. I've done it. My wife does it with us. We do it as a family. There are grandparents in it. There's actually a family in there with like three generations that are all participants in it. And everyone kind of brings their authentic self to these things. We sing all different kinds of music. And it's a very broad cross section of the community. And so a very welcoming one. And you know, in a funny way, it's fun. It's fun. It's not weird, but it's fun to do something you're not good at that pushes you in challenges you in ways that are very different than work. I enjoy that. Excellent. Well, when's the next performance? There's a lot of listeners in the Boston area. That's true. So I should know this off the top of my head. We're currently singing the Indigo Girls and REM, which I didn't know this before. We started both are from Georgia. And actually we're both their sort of brother sister bands. REM is slightly older. Indigo girls like younger, but like the manager or REM was the ones who discovered the Indigo Girls. And they've actually performed together as a couple songs that are written by REM. They were performed by Indigo Girls, etc. And the concert is in like three or four months, but I don't remember the exact date. But we're early in the season. Excellent. Great to chat with you. Yeah, likewise. Appreciate it. (upbeat music)