AI Meets IGG: How Indago Is Scaling Cold Case Solutions

Laura: Isaak and Steve, thank you so much for joining us today. We are so excited to have you at ISHI. Our 35th anniversary. Steve, welcome back. Good to see you again. Before we get into the amazing presentation that you’re doing, why don’t you each introduce yourselves and give us a little bit about your background?

Isaak: Sure. I’m Isaak. I’ve been doing AI machine learning for about 10 to 15 years. I lost count now, but that’s kind of my background. I’ve had the privilege of applying it in so many different industries. And so obviously, for the last about two years, I’ve been working in forensic genealogy with Indago. And I’m the CTO as well.

Steve: Excellent. Thanks for that, Isaak. Steve Bush. I’m the CEO of Indago. I spent my career at the FBI, as we talked last time, Laura. I resigned from the FBI in 2021 and so founded Indago with Steve Kramer and with Isaak Boseman, with the three of us are co-founders of the company and just excited to be back here at ISHI. Lots of fun stuff going on here. And thank you so much for the time here at the interview.

Laura: No, thank you for being part of the panel and taking time to do the video. We really appreciate it. And Isaak, it sounds like you have a very impressive background in AI and machine learning. What inspired you to apply that to forensics?

Isaak: It’s a good question. Yeah, I kind of got into it well before it was it was popular. It used to take me about 15 minutes to explain to people what I do. And now I just don’t say I’m an AI, you know, because, that sparks other conversations. But, I think, you know, more than anything, it was one of those I felt underserved areas because, you know, AI is being applied to every industry, and sometimes it’s applicable and sometimes it’s not. I felt like as soon as, you know, the Steves sort of showed me the process of how this works. I was like, wow, this is a really interesting area. So, I originally got involved more on the specifically the relationship prediction side of side of things. And so, it’s kind of grown out of that. And I think I still feel it’s an area that AI and machine learning is really applicable for. So yeah, just the opportunity was something I just couldn’t turn down, you know.

Laura: Well, I think that’s fantastic. I mean, I can only imagine how much more you can do, especially when we’re talking about IGG and applying machine learning and AI.

Isaak: Totally. Yeah, totally. For sure. It’s I mean, it’s ability to represent like so much data that, you know, like it just as humans, like we just don’t have the capital to go through all of it. It’s just able to process that, understand it, catalogue it, index it, make meaning out of it. You know, we’ve just not been able to do until now. So yeah, it’s exciting. Very excited.

Steve: I got to tell you Laura, it’s the biggest blessing in the world having Isaak being a part of our team, the good Lord brought him to us and it’s for a reason. And from the very beginning when we started doing these cases, it didn’t take long for us to realize all the stuff that Isaak just mentioned. We were we’re talking large amounts of data. You have to make connections between that data. You have to draw patterns within that data. And all of this starts, you know, the machine learning and the AI bells start going off. And anyone that’s been around Steve Kramer and I for long enough know that we’re not smart enough to build this type of software on our own, which is why we have to bring in the smart people like Isaak to do it. So, we’re just very blessed to have him and excited to have him on board.

Isaak: Thanks, Steve, I appreciate that. I think to add on to that, like I, I feel from the AI side, like it’s such a powerful tool, but it can also be quite dangerous, you know, and I think that what’s really important to me personally is to use it for good. And so as soon as I got involved with the Steves, you know, I could I could tell that this is something that’s going to not only benefit society, but, you know, their integrity. I felt lined up with mine where it’s like, we’ve got to make sure that we use this ethically and we use it in the right way. And so that’s kind of been, I think, a core value of our company from the beginning. And so, it’s just something I could get on board straight away.

Laura: Well, that’s nice to hear because I think, you know, that is a big conversation when you talk about AI and machine learning. So, it’s good. But before we even delve more into what you’re going to be talking about in your presentation, Steve, let’s talk a little bit about for people who haven’t seen our earlier interview or weren’t at the ISHI where you presented last time, you had a really interesting career trajectory from private sector FBI to working on your own. And what inspired that leap?

Steve: Well, that’s a that’s a story that’s probably longer than we can tell here. But yeah, I was I was in the private sector before the FBI as an engineer. So that was what I went to school for. I was a school trained engineer. The FBI won’t take you directly out of college, so I had to go have a real job before I could go to the FBI. But after I did five years in the private sector as an engineer, I started at the FBI. I think last time we talked, I was mainly a tactical guy at the FBI. I was a SWAT team leader and a firearms instructor. I was a sniper at the FBI, and those were really fun things to do. I really enjoyed doing them. But in in 2015, I was in a really big accident at the FBI. I shattered my spine and God had different plans for me. And, you know, I came back to the FBI after 21 months of physical therapy. And this was right at the time that Steve Kramer was just getting involved in the Golden State Killer.

He had assembled that team, you know, you know, Kurt Campbell and Monica and Barbara Rae-Venter, Paul Holes, Melissa Pariseau, who was the very gifted analyst at our FBI office in Orange County. The six of them solved that case. And that’s when we had this epiphany where we realized, even though the case solution itself was a huge win, it was the process that was actually the bigger win. It was the technique that was the bigger win. And that was where, you know, Steve Kramer and I got together and said, let’s just build this as big as we can and what’s crazy, Laura, since we last talked, that team at the FBI that started with just the two of us, it has over 200 trained people now, the FBI investigative genetic genealogy team, and they’ve solved over 175 cases, very difficult cases, like cases that other people have tried to solve and haven’t been able to solve. They bring those to the FBI, and 175 of those have been taken care of in of in the last six years, which is pretty cool.

Laura: That’s remarkable. I mean, that’s incredible to hear. So that that must feel good, you know, to watch that progression.

Steve: Well, there’s guys that are presenting here today from that team. Mark James is the agent from Baltimore that presented yesterday. John Hoffman, Stephanie Mellinger I mean, these are fantastic people. Amy Whitman was supposed to be here, but she had some other stuff come up. And that team of people lives on at the bureau, and they’re just doing a great job. And it’s cool to see that.

Laura: Well, that’s fantastic. Well, with both of you coming from diverse backgrounds and then bringing it to forensic genetics, I mean, we could talk about both the challenges and the opportunities that come from bringing those kind of backgrounds into something new.

Steve: I mean, I’ll tell you this. For me, it was actually a very logical flow because engineers like when you think about what engineers do is engineers are given problems. They’re given variables and they’re given formulas that govern like how those variables interact with each other. And then you have to piece them together, get to get to an answer. So, when I saw my first genetic genealogy case, I could not spell DNA. I didn’t know what a first cousin once removed was. Literally, I had no idea! We had to figure these things out. But that training, that engineering background actually made it very easy to transition into something like that. So, I found it to be enjoyable. And it was one of those things that we were able to pick up pretty quickly and start to get success fast. We were getting case solutions fast. And that was what was really exhilarating, because you’re solving cases that people say, you know, are unsolvable cases that have been solved, unsolved for ten, 20, 30, 40 plus years. And you’re coming in and, you know, having a case solution in just a just a few months or a few days even, it’s a big deal. And then you start talking about applying artificial intelligence to that mountain of unsolved cases. And that’s when it becomes a tool that changes the world.

Laura: Yeah. Let’s talk more about that, Isaak. Yeah, sure.

Isaak: Yeah. I mean, I think from a challenge perspective, so much of the skill is transferable. I think what’s super helpful for helpful for me is normally, you know, when you do an AI project, you’re really looking for subject matter experts, right? And so certainly my background, I didn’t have any real genealogy or genetics background, but fortunately I had two of the leading experts in the field. Right. So that’s great because, you know, then if I’m building the models that, you know, implementing the technology, I can get feedback from them straight away. Right? And so I think it’s been an absolutely, you know, easy from that perspective. Obviously, we’re kind of, you know, breaking new ground. So that’s always a challenge in itself. But I think other than that, you know, I feel like the diverse background has helped, you know, translate to this. Yeah.

Laura: I think that it’s always helpful to have that. I mean, it’s nice to have fresh eyes and then to have some experience and put that together and see what happens.

Isaak: Because I don’t have a law enforcement background at all. So, you know, it’s kind of great to sort of see how these guys do what they do. And then I’m a little bit on the outside looking in, going, I think we can we can make that better. We can speed that up, you know? And they can also tell me that, yeah, that’s not going to work, you know. Okay.

Laura: Well, I think what’s really interesting is how you’re going to apply this to IGG, FIGG, whatever we, you know, whichever term we want to use to degraded DNA samples, complex family trees. How is it going to impact how you’re able to work with that and how quickly?

Steve: Well, the interesting thing is the lab technology continues to get better. I mean, you’ve got a bunch of the labs that are here today. BODE is a great example. I was talking to the BODE folks earlier today. And they’re getting SNP profiles, the profile that we need for genealogy from lower and lower amounts of DNA. So lower quality, lower quantity samples are now producing SNPs that are usable for genealogy, which is where our process really starts. I kind of break it into the front end and the back end. The front end is going from, you know, DNA to a profile like from the crime scene to a profile, and then the back end is just what we do, which is from that match list that’s created from the profile to an actual identity. And so, it’s exciting for us when cases that had DNA samples that previously would not produce SNPs that were usable, they’re now producing SNPs that are usable that we can ingest into our software to make solutions. And so, yeah, it’s exciting.

Laura: That’s fantastic. Isaak, do you want to talk more about the software and the tools that you’ve developed?

Isaak: Yeah, sure. I mean, I think the real, if we can call it the secret sauce, is obviously the ability for a machine to learn how to build these trees and also to be able to correlate public record data to these relationships. And so, I think, you know, the two sides of that is obviously the algorithm that drives it, but also the access to the data. And so, we’ve been fortunate that our data and access to that has gotten better and better as we’ve gone through it. And so yeah, I think the great thing about and the difference with AI is, you know, normally if you program something like this, you know, you’re trying to, you know, take certain input and get guess or an output. And with AI it’s the opposite, right? So, we’ve got all these great cases that that’s been solved that we can actually use to teach the machine to do that. And I’m happy to dive into that. But I’m probably going to, you know, just tank the whole interview if I do that, you know.

Laura: No, I don’t think so. I mean, I think we want to hear about the tools and how you’re using them.

Isaak: I’ll let you take that.

Steve: One of the things that Isaak just said, that I think is so incredibly important when he talks about the data that we use, you know, it’s taken a while to find the right data providers for us to be able to partner with. Yeah, and we’ve got a great range of partners that we have engaged with, even since you and I last spoke. Laura, we’ve got a company founded out of Nashville, Tennessee, small little startup that we’ve been using. We’ve also got a Texas based company right here, not far from where we’re sitting. And then just recently really excited about LexisNexis. I mean, Lexis is the gold standard in data technology. And we’ve got an agreement with them, a partnership with them that’s been very, very fruitful for us. And, so just to kind of double down what you were saying, Isaak, you know, having quality data partners that can provide us with what we need is very helpful because without that, you can’t you can’t do the most fundamental part of, of genealogy, which is inferring genetic relationships from public records. That’s what all genealogists do. And that’s what you know, the machine learning focus really has been for Isaak.

Laura: Which is I think fascinating for our audience. So, with these partnerships, these amazing partnerships, strong partnerships that you have now, how are you taking those partnerships and turning it into the next step?

Isaak: Yeah, I mean, that’s a good question. And certainly I can get a little more technical on that side.

Laura: I think we’d love that.

Isaak: I think obviously the advent of AI as we know it today is really just a subfield of machine learning called deep learning, right? Which we’ll talk about a little bit more tomorrow, but it’s ability to ingest massive amounts of data and create these hyper dimensional correlations that we just can’t, you know, again, you know, talking about going through that manually. It’s like a really cumbersome process. So, what we can do with AI is really find those correlations much quicker and much faster. So really, we’re working on that side of it. So a lot of it’s obviously got to do with data processing. But more than anything, you know, giving the machine all these examples of relationships that’s inferred through Steve said through public records, you know, that we were kind of just automating that. And really at the end of it, it becomes like an automated identity resolution system. Right. And so that’s really how we’re, you know, and again, I can I can dive a little deeper if we need to.

Laura: Well, what I think is interesting for people and even on a high level, even if we’re not getting that technical, I mean, everybody knows genetic genealogy. It’s been in the news for years now with cold case after cold case being solved. But it is labor intensive. Sometimes it’s just individuals. I mean, many times there were people working over their coffee table and then, you know, taking out to the public. Now when you take AI and machine learning and apply that, I mean, I can’t imagine what that looks like as an advancement. So, I’d love to hear more.

Isaak: Yeah, totally. And I think what’s crucial and what makes deep learning so difficult is that it’s not very explainable. Right. So, I think even, you know, as a practitioner, there’s a lot that won’t admit this. But, you know, a lot of us just don’t know why it works. Right? It’s an empirical science for the most part, because we’ve got these, you know, neural networks that are just so vast. And so, we’re trying to find valuable connections within them to try and explain how is it able to make these decisions. And then we get scenarios like Tesla, you know, the car makes a decision to, you know, veer off the road and cause an accident. Tesla will go and try and create exactly the same conditions, and they just cannot replicate the AI to make the same decisions. The great thing with this application is we always have the genetics to fall back on. So, if it makes certain decisions about certain relationships, well, we can always go back and say, well look does the genetics really line up for what we’re trying to do. And so, I think that’s super valuable. But yeah, absolutely. I think, you know, anything that we can help from the genealogy perspective to help people make decisions faster, I think is super valuable. And the ideal thing is it just keeps getting better and better and better. Right?

Steve: So, yeah. Well, to tease that out a little bit more, one of the best parts about this process is we have an answer key that tells us whether we’re right or wrong. I tell people when I was an engineer, you know, engineers do a lot of problem sets. And my favorite part after doing a problem is to flip to the answer key and see if I got it right. Well, the answer key for genetic genealogy is that collection of a sample and the comparison of the suspect sample to the crime scene sample at the conclusion of this process, which tells us whether we’re right or wrong. And that’s that’s how we figure out at the end if the I everyone you know, I’ve heard people say that, well, how do you know that the AI is going to make the right decision? How do you know that genetic genealogy is even making the right decision? Well, the reason we know is because once we find the suspect, we take that suspect’s DNA. We compare it to the crime scene DNA. If it matches, we got it right. Then we got the problem right. If it doesn’t match, we got it wrong. And so, it’s virtually impossible to arrest the wrong person. If you take the time to do a surreptitious collection and confirmation of the DNA at the end of that process.

Laura: So, what it sounds like to me is if you’re looking at the Tesla example, rather than not understanding why it went, you have guardrails, you put guardrails in place and then you’re able to come to the right decision no matter what.

Steve: Yeah, I used that exact word all the time. No, it’s great. It’s fantastic. I used the word guardrails. The genetics act as a guardrail. So, when you when you’re autonomously constructing a pedigree, a family tree of a person, you’ve got DNA data points along there that act as guardrails. They say, look, you can’t go that way because the DNA won’t let you. You can’t go this way because the DNA won’t let you. That’s one part just for the actual construction of the trees themselves. But when you talk about finding the right guy, you have an even additional check at the end of that. That’s the, you know, the collection and confirmation of the suspect’s DNA. So, yeah, you have a guardrail on top of a guardrail, if you will.

Laura: Which I think gives a comfort level, because when, of course, when people just think AI, machine learning, there is that fear that you mentioned about it, you know, taking it the wrong direction. So that’s very a comforting way to think about it and look at it and exciting for what you’re doing. Do you want to talk about the sample, the Australian 14?

Steve: Yeah, it’s interesting. The Australian, the VIFM, the Victorian Institute of Forensic Medicine, some fantastic folks over there that we’re going to talk all about them tomorrow. They approached us and asked because, you know, Australia is trying to figure out how can we have success that the United States is having with this, with this same technique. But they have different considerations for data over there, different ways that they deal with public data. And so, they asked us to do an international pilot study to see, you know, if you if we took known Australian donors like people who have Australian connections and we ran them, we ran SNPs on those people and put those into American databases like GEDmatch, and then looked at those matches. Is there anything that our software can do to help figure out who those people are by using public records in the US? Now, interestingly enough, before we even did this study, anecdotally, we already knew this was going to work because Steve Kramer and I at the FBI, we did plenty of cases where, you know, you had a… Here’s a great example. One of the first cases we did, the homicide of a of a mom named Barbara Becker in La Jolla, California. We started doing the genealogy on that. All the people were in Canada. And so, we did a bunch of Canadian genealogy and then found out that there was a Canadian who was adopted to a US based family.

He was the one that committed the crime. So international nexus to solve a US based crime. We did another one out of Riverside where the genealogy was all Swedish. We were using Facebook and, you know, Swedish records to find a Swedish person who immigrated to the US. So, this question was, can we do it backward? Can we take people from Australia and see if it worked and it worked. You know, we knew it would. You put you put those folks’ samples into GEDmatch. You’re able to look at their match lists, and then we’re able to pull public records on those folks from the US. Because keep in mind, if you have a third cousin, you know, your most recent common ancestor is going to be 3 or 4 generations back, and that that ancestor might have had a line of people that went down to Australia and a line of people that came down to Texas. Right. And so, we’re able to find records on those folks. And so, I think for them, it built some confidence that they could use this system and help identify suspects. And, you know, unidentified human remains from Australia with the process.

Laura: What a fantastic way to illustrate, you know, the power of it and where you’re going next. Do you want to talk a little bit more about that? Like what does this look like for you guys? I mean, I’d love to, you know, what do you predict when we when we speak again next year or the year after that, what will we be looking at? We’re best guessing here. It’s just a thought experiment.

Isaak: Minority report.

Laura: Oh no.

Isaak: Obviously, you know, as we’ve discussed before, our data sources keep getting better. And so, with this pilot study, you know, one of the next steps we want to do is apply our new data sources to that. So, because obviously we were quite limited in the public records that we could use when we’re trying to build the pedigree. So that’s certainly something that we want to do. And obviously, you know, AI is an iterative process. So, the goal is that we keep getting better. You know, certainly, you know, the great thing is we can do that retroactively as well for any other cases. So, we kind of didn’t really talk about like sort of going a step extra where we can rank some of these cases and we can, you know, sort of figure out and prioritize some of them. But then we can, you know, as we solve some of these, we can actually retroactively run through them as well. And they should, in theory, improve our pedigrees that we already have.

Laura: I think the retroactive. That’s very interesting. I’m glad you brought that up. Yeah.

Steve: I think, Laura, when you look at the big picture of what’s happening in DNA cases here in the United States, CODIS is a good place to start. Tom Callahan is here. I think he presented too. Tom’s great. He’s been a guy that we’ve worked with a lot on this. The father of CODIS, some call him. But when you look at CODIS and this isn’t a knock on CODIS, I’m actually a fan of CODIS. And CODIS does a fantastic job. But when you look at CODIS, CODIS will tell you if you go to fbi.gov right now and you look it up, it’ll tell you that there are about 1.3 million unidentified forensic profiles in CODIS. Those are all unsolved cases. And if you look from 2018 until now, what that delta looks like as it goes up, it’s going up by an average of about 80,000 unidentified profiles per year. Genetic genealogy has only solved about 1000 cases in six years is what the estimate is right now. And so, it’s not really making a huge dent. And don’t get me wrong, those 1000 cases are very, very meaningful. But when you have a backlog of over 1.3 million, you know, going up by 80,000 a year, it’s not really making a huge impact. And the reason for that is there’s no software to be able to scale this. And that’s where, you know, Isaak and I really have and Steve Kramer, we really just have our hearts invested in that because we realize providing that tool to the law enforcement world that solves cases well beyond our years. Like, well, after, you know, Isaak and I are retired and dead and gone and the next generation of police are using that tool, it’s going to continue to solve cases into the future. And that’s what gets exciting is really, you know, really changing the game beyond just what, you know, this small little sample set is that we’ve looked at thus far.

Laura: I think that’s what’s very exciting about your company is scaling it via tool so that everyone. Yeah. You’re right. I mean, you know, working one by one, there’s just no way you’re going to address all of those cases. Yeah. What have we missed? What else do you want to tell us about your presentation tomorrow?

Steve: I mean, the presentation tomorrow is it’s very small. It’s only 20 minutes, but great team effort. I mean, it’s so cool. Like, I was joking with Isaak earlier. I asked what happens when you put, you know, nine people from two continents together? Four doctors, an engineer, a data scientist, a lawyer, a little Vegemite, and a little apple pie. You know, it’s like it came out.

Laura: It sounds like a joke.

Speaker4: Yeah, exactly.

Steve: Sounds like the beginning of a joke. But. But it was a really cool study, just a really cool team effort. And, you know, the international collaboration I think is great, especially for countries that are looking to the US to say like, hey, here’s an example of the right way to do it. And, you know, we’ve had lots of conversations with other places overseas, Europe, Canada. And it’s like they all are looking to the United States for leadership. They’re looking to the FBI for leadership, US law enforcement. They’re looking to us for leadership. And so that part makes it exciting. And just the team effort, I think, is what I’ll take away from that.

Laura: Yeah, absolutely. Isaak, anything you want to add.

Isaak: No, I think I think I can’t say it any better.

Laura: That’s very impressive. Well, I can’t wait to see the presentation. This is very exciting. We’re very lucky to have you here. And I can’t wait to see what happens next as well. As long as we have you, Steve, I know you’ve been here before. And in honor of our 35th anniversary, we’re asking everyone, you know, how do you like ISHI? What do you like about it? What else can we do for you? And Isaak? Your first year, I believe? Yes. Okay, so as it’s Wednesday, we can also ask you. What do you think so far? Yeah.

Isaak: Yeah. I mean it’s been great. Like, obviously it’s a fairly new community to me. I think obviously, I feel like, you know, it’s a community that I can really contribute to. So that’s always nice, you know, and I think everyone I’ve met so far has been fantastic. Really smart people, you know, and it’s always a great opportunity to learn, you know. So it’s been great, I think it’s a fantastic conference.

Laura: We’re happy to have you here Steve.

Steve: Well, you guys continue to up the ante on the venue. This venue is absolutely amazing. I don’t think you could pick a nicer place to host something like this. When we walked in, I was overwhelmed, like, wow, this is a really, really a nice place. But I think, I mean, the big, big advantage to anybody that’s looking to come to ISHI is the networking. It’s the networking opportunities that it provides. And there’s so many great presentations, you can’t see them all. It’s not even possible, but it’s the downtime between the presentations. It’s being in the lunch line. You know, it’s walking through the exhibit hall and hall and getting to, you know, bumping into people and rubbing elbows with people, putting names with faces. Certainly the after evening stuff. I mean, sometimes there’s some people down there pretty late last night. You know, talking to each other. That’s where you really where you really connect, where you really add value, you know, to what happens here. So ISHI just continues to be, you know, a very rare place to do that, because we go to a lot of conferences and none of them are really like this one.

Laura: Well, we appreciate that. I agree, anywhere you can go and have that in-person connection, it makes all the difference. So, we’re lucky to have you here to be part of that. So, thank you.

Isaak: Thanks.

Steve: Thank you. We’re very appreciative of the opportunity.

Laura: We appreciate it. Thank you.