Shyam Sankar
speaker
319 appearances
2 recordings
2 series
first heard Apr 2025
last heard 6 May
Shyam Sankar’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in May 2026 with 1.
Appearances
It's shocking how hard this is. How do I have access to all of the data that's in my enterprise in a way that I as a human think about it? I as the principal, not just the human, but I as the principal, like what's my model for this? What are the questions I want to ask? Simple example, when Rocket Man was rattling his saber in 2017,
In North Korea, the Army wanted to answer the question, how many tanks are there in the Army? That is a three-week data call. Are you serious? It takes three weeks to get an inventory? Until we got involved, it took three weeks to answer that question. Because there is no canonical representation of a tank. And what do you mean by tank? Do you mean the ones that are ready or not ready?
So the way you would do it is you would send down the data call to all of the units. They'd go check out the motor pool. They'd come back with the answer. And so there's no living, breathing, canonical record that continues to flow through this. We've built these systems. The army is older than the country. We've built these systems.
It's like archeology when you go back there and try to look at this. It's not how you would do it today, but that's our extant reality. We have to deal with this messiness. Sometimes the cynical way to think about Palantir is it took something as sexy as James Bond to motivate engineers to work on a problem as boring as data integration. But that is the starting point.
If you can't see yourself, if you can't integrate all your data, you're just always chasing your tail. Then on top of that, we make powerful interfaces for decision-making. One of the contrarian quips I have is that Data is not the new oil. People have been running around saying, data is the new oil, data is the new oil. I think data is the new snake oil.
There's nothing inherently valuable about data. It's only valuable if you can use it to make a decision. So it's about decision advantage. So how do I leverage this data I've now integrated so that you can make a better decision, you can see further in the future, that you can out-compete your competitor in the commercial world or your adversary in the defense world?
The decision is leading up to the point where you realize that's what they're trying to do. How do I integrate the human reporting that I have that tells me what might be going on? Now, keep in mind that... One part of the reporting is in this country. One part of the reporting is in that country. Who's willing to share what?
Who even knows that sharing it is going to lead to a mission critical outcome that saves lives? So how do you start to automate more of that? How do you help them piece those things together? How do you combine the technical collection you have with this too?
Maybe you have SIGINT that helps you understand this, helps you understand that this reporting is correlated to other cells that you actually care about, where you have intercepts that tell you something.
How do I use my historic FMV footage, observations of that compound that helped me piece together more things that lead to the conclusion, we've got a problem here and we've got a tight timeline to act on.
It's like shining, it's like turning on the light on the battle space, the things that you couldn't see before you could see now. Now, some important, might seem slightly technocratic, but we're providing the software to the government customers. No one's providing us all of this data. The government has the data, great, but it's just sitting there. It's in the dark.
It's hard to see all of this information. It's overwhelming. Great. How do we get the spotlight to highlight the things that actually matter? How do I get to ask the next question? Here's something that's risky. Okay, what are the next 10 questions you're going to want? Can you even answer those 10 questions in 10 seconds, or is that going to take you 10 weeks?
If it takes you 10 weeks, you're not even going to bother answering those questions. So can I make that fast enough that you get to why does this matter and is this a threat or is this irrelevant? That's the first part. The second part of it that's really important is because why is this information sharing so hard? Well, there are lots of rules and regulations. People have different authorities.
Different things can be shared. The way we enforce that today is with humans. which is crazy, which is why it's slow and inefficient and you miss things. We replace that enforcement with software. The software ensures no one can see anything they're not allowed to see.
The software ensures under the right conditions for information sharing, the right pieces of information flow from one agency to the other. So by automating that flow, it means that you kind of have a hive mind.
the entire government can operate competently because you're actually able to see everything you're allowed to see, as opposed to, well, we have humans who are gatekeeping us along the way, just slowing everything down, which always, when you have humans, it always devolves into control.
The mission gets obfuscated by, well, we do this job, they do that job, and the interpersonal factors get in the way.
Well, you know, it was really hard because none of us had worked in government. None of us had clearances. We would go to D.C. We'd literally carry our Pelican case with a projector. I mean, talk about state of technology. You couldn't even rely on a projector being in the government conference room.
You had to bring your own projector to make sure you could actually show the customers what you were building. And we were really eliciting feedback from them. Like, okay, I built this. And we don't know your workflow, but based on your reactions, I'm going to go away and code that night and come back tomorrow and show you something new.
And we had to do all this kind of notionally to begin with on low side synthetic data. until we got to a threshold of conviction. And that threshold of conviction, I mean, almost to the point, it started because of a renegade analyst. And I think she's been a prior guest of yours, actually, Sarah Adams. No kidding. Yeah, so Sarah Adams.
Showing 41–60 of 319 · page 3 of 16
← Previous
Next →