Enterprise Scale AI with Private Data – Agents – Six Five On the Road

Enterprise Scale AI with Private Data - Agents - Six Five On the Road

On this episode of the Six Five On the Road, hosts Alastair Cooke and Lisa Martin are joined by‘s Luke Norris, CEO for a conversation on how is leading the charge in the AI revolution, aiming to redefine the enterprise scale AI with a focus on private data agents and autonomous processes, enabling the 5th Industrial revolution.

Their discussion covers:

  • The mission and vision behind, aiming for 1 trillion inferences per day.
  • How data privacy and efficient processing stand at the core of’s technology.
  • The use of ‘Inference Mesh’ and RAG technology to deploy AI across various environments.
  • The significant role of agents and autonomous processes in driving the 5th Industrial Revolution.
  • The future of enterprise-scale AI and how is setting the stage for new possibilities in the industry.

Learn more at and

Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.

Or listen to the audio here:

Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.


Lisa Martin: Hello, everyone. Welcome to Six Five On the Road sponsored by Solidigm. I’m Lisa Martin and I’m very pleased to be joined for the first time, IRL, in real life with Alastair Cooke. It’s so nice to finally meet you and host with you.

Alastair Cooke: Thank you, Lisa. It’s great to be here. I’ve been co-hosting on many things amongst the wider group, but it’s great to be here on the Six Five On the Road.

Lisa Martin: Absolutely. We have a great guest up next. We’re going to be talking with Luke Norris, the CEO at Kamiwaza. And I did a little LinkedIn stalking of Luke and he describes himself as the wearer of white shoes, which I can verify he is wearing white shoes, but also the builder of companies that make an impact. Luke, it’s great to have you on Six Five On The Road. Love the white shoes. You are on brand. I’d love to know a little bit about Kamiwaza. I know the name means superhuman, but help the audience understand what you’re doing. I know you’re industrializing reasoning enabling the Fifth Industrial Revolution, making it easier, but that seems like wow, how do you do that?

Luke Norris: Well, thanks. So we’re really focused on making AI operational and effective at scale for the enterprise. Right now, everyone seems to be focused on sort of training models, which makes little sense. It’s the big numbers, it’s the big processes that are happening, but when you look at what the enterprise is actually doing, they’re not training models, they’re trying to actually utilize models, affect the way that their operations and services work. So we’re building a full stack solution, real turnkey capability, so the enterprise can run private models on private data from cloud, edge and core.

Lisa Martin: And talking about that, what are some of the widespread adoption? We talk about gen AI a lot, the influx of momentum and acceleration that ChatGPT brought to the world 18 months or so ago. But talk about some of the advanced technologies that you’re seeing organizations adopt that you’re helping to make easier.

Luke Norris: Yeah, it’s amazing. Everybody sort of had this ChatGPT moment and they think of gen AI as a chat interface. And it’s literally couldn’t be worse than that. The actual sort of evolution of it is an autonomous system, typically called agents, and these agents actually can be given a task. They can be given data and they can operate 24 hours a day, seven days a week, literally at superhuman scale and capability. A great use case for a Fortune 500, one of the largest logistic companies in the world. They have over 980 million PDFs, just shy of a billion PDFs.

Every contract, entitlement, invoice, everything that this very large company’s done. GenAI has been able to ingest all of those, understand all their entitlements, put it into multiple formats, put those formats into systems like Salesforce, systems like ServiceNow, and move it back and forth. And now they have sort of an oracle of all their purchasing capabilities and it’s running there 24/7. As new service orders land on that file system, it ingests, process, redoes that whole transaction and it’s there for you.

Alastair Cooke: And Luke, it strikes me that that’s an awful lot of work that if you’re building this from the beginning, that will be months if not years of a team of a dozen, two dozen people developing this. Is that the special sauce that Kamiwaza is bringing in is to take away that requirement for huge numbers of people doing huge amounts of science in every single customer?

Luke Norris: Yeah, that’s correct. It was actually a hundred people on classic ML AI that was trying to get deterministic, read this, OCR this and actually understand it and make it. What we brought to it was the ability to have a full gen AI stack so that you could do the embeddings, you could do their full RAG process, you could actually then run multiple models in that agent structure. Also, the data was partly in the cloud, it was partly in multiple data centers, and it’s also at the edge where the actual people are scanning the files in and loading it up to local file stores. So the ability to have one stack across it all that had the same models in a large inference mesh, made this work. I kid you not, the very first output took 15 minutes.

Lisa Martin: Wow, that’s substantially fast.

Luke Norris: Yeah. Yeah. Well, I mean, that’s gen AI. So it was a plethora models. It was a couple mixture of expert models, a couple of smaller models, a model that did OCR, sort of optical character recognition and then actually re-looked at it and said, “Look, this part didn’t come through right.” So then it applied natural language to actually fix those characters, but that’s literally what the models were actually built for. And once you actually just input the data with an output sort of facilitated, you wanted it in this schema on the output, it was a very quick process.

Lisa Martin: That’s awesome. So I was reading a bit about Kamiwaza.AI, and you guys aren’t just part of the AI revolution. You say, “Hey, we’re leading this.” That’s a very confident statement. But we talked about your mission being really to empower organizations, enterprises to get ready for the Fifth Industrial Revolution, which is we’re on the precipice of, and you have some big goals here, and I saw aiming to achieve the unprecedented scale of 1 trillion inferences per day.

Luke Norris: That’s correct, yeah.

Lisa Martin: How? And talk about the impact, the potential impact that has.

Luke Norris: So roughly a trillion inferences a day, even with the latest GPT-4o, we roughly think would be about $3 billion a day in costs. But when we look at these very large organizations in the Fortune 500, for 20 to 30% of their operations to be fully automated with gen AI, that’s roughly what it’s going to take is about a trillion inferences. A full agent process, like I talked about earlier, might be about six to 10,000 inference requests to actually run that whole pipeline and have that output and go back and forth. So to get to that level of cost, we’re already working with many suppliers such as like Qualcomm and Ampere that are going to help drive down the cost of inference. And I can accurately predict over a thousand X in the next six to nine months.

Lisa Martin: Wow.

Alastair Cooke: How ready are organizations for this kind of volume of inference, volume of data transfer, in particular? These files that we’re talking about as these vast numbers of PDFs are typically stored on low-cost storage that we hope we never have to read back from. But clearly to train from them, we’re going to need something that’s going to help us to access that data. Are customers ready?

Luke Norris: I think there’s two folds to that question. I think the best thing that came out of COVID is the fact that all these organizations had to digitize very quick. We’re talking a 10-year compression happened in two to three years there. I mean, just absolutely amazing. So for a lot of these organizations, they do have the data in a place, it’s probably not the right place and it’s probably not on the fast drives. What actually happens in these large agent processes, like I talked about, is the slowest piece of it is the ingestion of the data in the third or fourth step of RAG.

And you’re not ingesting just one file, it’s typically tens, hundreds of files because the AI system doesn’t know exactly which one was the right one. So it keeps adjusting, re-ranking, adjusting, re-ranking. The whole process is slowed down by the data ingestion. So if you want to speed that up, you want to have that data local for latency, and then you want to have it on the fastest drives that can actually input it into the model as fast as possible.

Lisa Martin: What are some of the key use cases that you guys are working on now that excite you the most from an impact perspective or from a humanity perspective, for example?

Luke Norris: So just at a high level on humanity. I think this Fifth Industrial Revolution, just like every pre sort of revolution, it’s happening so much faster than we actually realize. And as these enterprises move to sort of having 20, 30, 40% of gen AI actually adopt theirs, I think there’s going to be a massive influx in sort of what a standard job is, what a standard role is within a company. But on that other side, is going to be finally where humanity starts to get to abundance. It’s going to be so much cheaper to produce goods, it’s going to be so much faster to produce innovation that I’m just trying to push the whole industry and the whole world to that level just that much faster.

Lisa Martin: And where is Kamiwaza on the… Obviously, we talk about all these things as journeys, the enterprise, the journeys that they have to go through. But talking about what you guys are aiming to do and really redefine the boundaries of what AI can accomplish, where from your seat is Kamiwaza on that journey to accomplish that mission?

Luke Norris: So I think we as a company, we’re at the precipice. We’re ready. We’re already having great outcomes with very large enterprises. What’s amazing about this technology, gen AI, this Fifth Industrial Revolution is it’s the large enterprise that’s the early adopter. They have the resources and they actually have the FOMO of not missing out on this journey. They know if one logistics company, if the other logistics company of similar size adopts it quicker, they’re going to be off to the races at an uncatchable speed. So these enterprises are putting the resources in to make that adoption. And I think the enterprise people were saying it’s going to be 18 to 24 months. I’d say let’s compress that to six to nine months because the boardrooms are putting the investments in place to do this now.

Lisa Martin: And I’m just going to ask you, where is that conversation happening? But you just mentioned the board. So this is C-suite level, this is board level imperatives?

Luke Norris: Yeah, this is posted on 10Ks to the public market, board-driven initiatives. They’re letting the people know that they’re investing into it and it’s trickling down to these companies extremely quick.

Alastair Cooke: And a lot of the value proposition for Kamiwaza is to be even faster at that and not be the six months, but getting down to the two and three months towards delivering value? This is-

Luke Norris: So it’s two-fold. We have an opinionated stack, so it works through a Docker image right out of the gate, all 25 packages from open source and sort of private enterprise technologies that we’ve built to deploy that same package, cloud, prem, and edge. And that really does help speed it up, instead of them making all the selection processes and vetting processes associated with that. Third is we’ve branched out past just NVIDIA to Qualcomm and Intel and so many other vendors to lower that inferencing cost, make that technology work across the stack. So as these new features and functions come out, the enterprises can adopt them in their natural process. And then the bottleneck is just on the business process. How fast can they adapt? And back to the boardroom, when they’re driving it, the business is adopting it quickly.

Lisa Martin: Talk a little bit about the partner ecosystem that you’ve built there. You mentioned a number of partner names already, but obviously the depth and the power of the partner ecosystem is critical. Enterprises want that partnership, but just give us kind of a snapshot of your overall partner strategy.

Luke Norris: So we’re really focused on, as I said earlier, driving the cost down a thousand X in the next six to nine months and then well beyond that. And the innovation has so far been just on that training-centric side of gen AI, and it’s just now transferring to the inference side. Literally, the graphics cards and the processors were built on technologies for video games. They’re finally now coming out with very specific processors and capabilities for inferencing. So we’re trying to expand that greatly. The other thing that we’re trying to work with is all the enterprise storage providers and enterprise services because you have to have this mass amounts of data, we’re talking petabytes and exabytes of data, interact with LLMs and gen AI in a net new way. It’s not just a data pipeline anymore, it’s actually a data notification. And what I mean by that is new data lands, data changes, it has to notify gen AI so it can reprocess, be aware of the latest updates of that data so that its outcomes are up to speed for the enterprise.

Alastair Cooke: And that’s the important thing about the retrieval augmented generation, the RAG methodologies is you can in a real time update streaming data and have that decision informed by much more up-to-date data than you can by doing the fine-tuning of a larger model.

Luke Norris: Yeah, absolutely fine-tuning processes, even continuous fine-tuning processes, which is sort of cutting edge art, still has a massive lag. It also, actually, a lot of new research is saying that that can actually cause unintended hallucinations because you’re changing all these weights and stuff that we truly don’t understand in these incredibly large models.

But with RAG, you’re feeding it the actual data for it to infer off of, therefore, it’s the precise lingua franca of an enterprise of actual data that they’ve looked at. So the output is directly reference-able to that, therefore you actually get the result. And that, when you work with enterprise storage vendors to make sure that you have the full data access, the full speed, and that pipeline, you can literally achieve that in real time.

Alastair Cooke: And it’s that reference-ability that makes this much more applicable to more regulated, more controlled enterprise environments rather than just creating a chatbot where the answer is disposed of nearly as soon as it’s given.

Luke Norris: Yeah, once again, in an agent workflow, you get to 80, 90% accuracy of mass amounts of data. There’s no way a single human would be anywhere near that. I mean, this is truly superhuman capabilities, hundreds of millions of data inputs, and it’s 90% accurate on what it actually processes. It will change all process and procedures of an enterprise. It is truly sort of the game changer.

Lisa Martin: I can see why you chose the name Kamiwaza, superhuman.

Luke Norris: Absolutely.

Lisa Martin: You talked about regulated industries. I’m wondering in our final minutes here or so, are you seeing regulated industries as more of early adopters? Obviously, every industry is going to be taking this up, but what are you seeing there in terms of any verticals or industries that are really on the fast track?

Luke Norris: Yeah, I think it’s counterintuitive. The regulated industry is the fastest adopter. They actually have the guardrails already set up in all of their process and procedures, which are easy to map to gen AI services and solutions. And that process and procedure that was baked inherently, the data processes that were baked inherently are very easy to port over to gen AI. Also, the regulators want that consistency. Gen AI is going to give you that 90, 98, 99% consistency of a process, like I said, a human can’t do at scale and speed.

Lisa Martin: Where do you want to point our audience to go learn more about Kamiwaza and what you guys are doing and the impact that you’re clearly already making?

Luke Norris: I would just look at Kamiwaza.AI, obviously our website, but we’re active across all social media channels, especially LinkedIn. And obviously you guys do a great job and we follow you.

Lisa Martin: Yes. Excellent. Luke, it’s been great to have you on the program. The wearer of white shoes, the builder of companies that clearly make an impact. I think we broke both of those down pretty well today. Thank you for joining Alastair and me on the program.

Luke Norris: And thank you too.

Lisa Martin: We appreciate it. For our guests and for Alastair Cooke, I’m Lisa Martin. You’re watching Six Five On The Road, sponsored by Solidigm.

Author Information

Lisa Martin is a Silicon Valley-based technology correspondent that has been covering technologies like enterprise iPaaS, integration, automation, infrastructure, cloud, storage, and more for nearly 20 years. She has interviewed nearly 1,000 tech executives, like Michael Dell and Pat Gelsinger, on camera over many years as a correspondent.

Alastair has made a twenty-year career out of helping people understand complex IT infrastructure and how to build solutions that fulfil business needs. Much of his career has included teaching official training courses for vendors, including HPE, VMware, and AWS. Alastair has written hundreds of analyst articles and papers exploring products and topics around on-premises infrastructure and virtualization and getting the most out of public cloud and hybrid infrastructure. Alastair has also been involved in community-driven, practitioner-led education through the vBrownBag podcast and the vBrownBag TechTalks.


Latest Insights:

Nivas Iyer, Sr. Principal Product Manager at Dell Technologies, joins Paul Nashawaty to discuss the transition from VMs to Kubernetes and the strategies to overcome emerging data storage challenges in modern IT infrastructures.
Shimon Ben David, CTO at WEKA, joins Dave Nicholson and Alastair Cooke to share his insights on how WEKA's innovative solutions, particularly the WEKApod Data Platform Appliance, are revolutionizing storage for AI workloads, setting a new benchmark for performance and efficiency.
The Futurum Group team assesses how the global impact of the recent CrowdStrike IT outage has underscored the critical dependency of various sectors on cybersecurity services, and how this incident highlights the vulnerabilities in digital infrastructure and emphasizes the necessity for robust cybersecurity measures and resilient deployment processes to prevent widespread disruptions in the future.
On this episode of The Six Five Webcast, hosts Patrick Moorhead and Daniel Newman discuss CrowdStrike Global meltdown, Meta won't do GAI in EU or Brazil, HP Imagine AI 2024, TSMC Q2FY24 earnings, AMD Zen 5 Tech Day, Apple using YouTube to train its models, and NVIDIA announces Mistral NeMo 12B NIM.