Building Remote Teams That Scale: Jim Olsen’s Leadership Playbook

Narrator: Welcome to Software Leaders Uncensored with Steve Taplin.

Steve Taplin: Welcome to Software Leaders Uncensored, where tech leaders drop the filters and just tell it like it is. I'm your host, Steve Taplin, serial entrepreneur with 30 ventures under my belt, where I failed at 20 of those and scale 10 of those to success and turn that roller coaster of a ride into my book, Fail Hard, Win Big. This book just went live here in July of 2025. Key ingredient I talk about is how to use failure as a superpower and how custom software has really been the difference maker in every win that I have done.

I'm now the CEO of Sonatafy Technology. We are nearshore software development company helping others build high impact software without the corporate BS. Today we are joined by Jim Olsen. Jim is a seasoned principal level developer, software architect, and hands-on CTO who's got some good stories.

So, everyone, buckle up. Let's have some fun. Jim, welcome to the show.

Jim Olsen: Thank you. Look forward to it, and thank you for having me on.

Steve Taplin: You bet. Jim, please uh walk our listeners through the journey of your career.

Jim Olsen: Okay. Yeah. Well, kind of started out unique. Uh I taught myself at nine to program.

My dad worked for Bell Laboratories and I had access to thermal terminals. So, you know, you leave the paper out in the sun, it turns black. So, I started out way back then, uh, and got my first job when I was 13. Um, and then, uh, did follow through and go to college, out of college, started with a startup in the open software foundation, uh, that we were working with there.

So, I kind of got my roots in open source software uh, which is was important to my career and understanding back early before I was really popular. Then because of that I was brought on uh into novel where I rose to a distinguished engineer which was a board-elected position on that where I was working on a lot of the standard protocols you deal with today things like LDAP internet printing protocol I mean all basic stuff now but back then didn't exist so we had to build it on that and through that uh I got involved with Netscape basically we did a joint venture between Novell and Netscape um I personally moved all of their software from their platform to Novell's platform and it resulted in my first kind of entrepreneurial effort which was a spin-off called Novonyx which was a combination uh joint venture between the two where they invested in it but we were an independent startup so that was as the chief architect of that as my first startup and got my feet wet there and learned a lot about how things you can be successful but still go wrong you know we actually got got our first product out the door but then both decided it was too important and immediately shut us down and took took all of our software we made so I learned you can do everything right and still from the technical side, but there's a lot more to running a company than just the tech side. So, it was a good first experience there. But then I actually ended up reporting directly to the CEO of Novell, who was Eric Schmidt at the time.

Um, so I got a lot of advice early on from him from his experience at Sun, etc. on that kind of things. And then, uh, did you know some varying startups as like you said, most startups do fail so most of those failed on that. But then I also purposely worked in some big big uh mortgage company to learn about the financial space and what that software development is like because it's entirely different than a Silicon Valley style company and learning what really goes on with the enterprise.

I moved into Qualtrics for a while and worked with them if you're familiar with them. I kind of they were a unicorn startup originally got acquired. I left before then to get into big data where we actually I was doing consulting and leading the consulting organization uh around big data efforts in very large companies. I was working with the likes of like you know Western Digital and HGST and uh uh you know everything across manufacturers to banks to etc.

So again I got exposed to a lot of what was going on and one of the things I saw is you know the whole reason for having big data was the fact that uh you know you needed to do something with it. You don't just collect data there's not much value in just landing data. So I was seeing data science efforts coming on fairly early on and that's what kind of led to this current company in that there was a lack of actually governance of all of that machine learning software at the time and of course that's led on to generative AI and now agentic AI and everything not having a process in place it most places were like the my early days in the 80s and 90s coding where yeah on my desk ship it to production we push it right out there. That's what we were seeing with machine learning where literally one person was the only person who could push a model that was literally being used by hundreds of machines and nobody else knew how to actually replicate it, roll it back or you know that person gets hit by a bus, god forbid, they were in trouble.

Um, and I saw this again and again and that kind of g spurred the idea of not only going down the CI/CD route where you know you have a process for how you get it out there but also gaining knowledge about what you're putting it out there cuz machine learning generative AI are non-deterministic in nature. Uh so you do need to have more understanding of a good process to build trust and that's kind of what led to my current role. I was actually brought as VP of engineering to uh to uh build the process but I kind of redesigned the what they had as a start and uh was moved into CTO role instead on that and we brought a full-time VP to more manage the people.

Steve Taplin: Well, so wow Jim you have done a lot in the development space both startups both large companies big and small. Um, and you live half the year in Park City, Utah, and half the year where you are now at a cabin in 10,000 square feet in Colorado, right?

Jim Olsen: No, the cabin's only 600 square feet, but I'm 10,000 ft high.

Steve Taplin: 10,000 I'm sorry, 10,000 ft elevation.

Jim Olsen: Yeah, my bad.

Steve Taplin: No worries.

Jim Olsen: But yeah, no, it's great.

Steve Taplin: Is that your way of escaping and being able to just focus without the distractions?

Jim Olsen: Well, it's it's a little bit of that. a CTO you work long hours you know if you're able to basically at the end of the day walk out go walk in the woods and relax and decompress it gives you time to uh kind of think about things differently and step back from the problems for a minute so it provides that peace and isolation and I love nature that you can actually uh do that kind of things or go for a ride on the uh UTV or whatever but it provides ready access to kind of an instant 2-hour vacation and then you get back to

Steve Taplin: And thanks to Elon Musk and Starlink, you can get good connectivity really being off the grid.

Jim Olsen: Yeah, so far it's working pretty good. I did one of the first uh actual DC conversions of his unit because I run on pure DC. Uh so we kind of hacked it and figured out the wiring to actually uh be able to run it off of DC.

Steve Taplin: Very nice. Now, so ModelOp, tell us about your company and what you guys are doing.

Jim Olsen: Sure, absolutely. the uh you know one of the big problems with uh the rapid rise in machine learning AI generative AI now agentic AI is understanding basically what is deployed where what went into it ensuring it's behaving as it's supposed to uh understanding the ROI of what like okay I'm using generative AI it's costing me how much in tokens and how much am I getting back in business on it who approved it and because the very nature You know, you hear about the hallucinations and you hear about the bad effects of like McDonald's, they deployed that uh AI solution in their kiosk and they got all these Tik Tok videos of people, oh, I ordered I'd like one fry. Okay, I added 10 fries to your things. No, no, no.

I only wanted one. Okay, I added another 10 fries. You don't want that happening to your business. It makes you look bad.

It can hurt your brand as well as cost you real money. So how do you actually build trust in AI applications to understand that not only uh are they behaving properly once deployed but how did they get there? Um and the foundation of that our first clients were things like banks etc who are under Dodd-Frank Act where they have to prove this stuff or they get fined they get audited they have to show they've done the due diligence to make sure it's making fair financial decisions for people because you can destroy people's lives with this. Well, now we're seeing where generative AI and other solutions are coming into things like health care where, you know, it literally is making life or death decisions about people and stuff and do you want to just trust the solution and, you know, not be able to defend that's behaving correctly or more importantly hurt your brand, hurt your company and, you know, and hurt real lives.

So you need a process for actually getting that out in place, understanding, building trust, understanding how it's performing and tying it back to a use case, not just a technology. So our software actually answers that process end to end from use case being we don't even build anything yet all the way out to uh you know monitoring and retirement and replacement and etc. as well. And we're really unique that way in that we're not a data science workbench.

But we work with all of them and instead what we do is we actually uh manage that overall governance process in an agnostic way.

Steve Taplin: Gotcha. So, you know, AI is obviously the hot topic over the past two years that at this point things are changing on a weekly basis and evolving and I mean for you said it best, but for most companies to take true advantage of AI, you have to have good data and then not only do you have to have good data and centralized data from everything, but then you need governance on that data or it's no good. And that's, you know, I think, you know, the next five years, most large corporations, that's probably one of their number one goals to ensure they can maximize AI.

Jim Olsen: Yeah, absolutely. I mean, that's some of our biggest customers are things like uh Fidelity, Procter & Gamble, etc. I mean, they're they're big companies that we work with. We're unusual as a startup that way.

uh you know there like they obviously have the largest need cuz uh you know you can't make wrong moves in those environments or you know there's real fines you know I forget which bank it was but they were fined like billions of dollars for making a mistake uh on that kind of things for having a bad AI model you can't do this stuff it's not okay never mind even if you're not a fine okay well now I hurt my brand reputation many companies brand is their most important thing like you know you're you're buying one brand of toilet paper versus another beyond quality issues. You know, brand makes makes an influence there. So, you absolutely want to make sure these things are performing correctly and it's understood and you also understand if it's worth the investment.

Steve Taplin: OpenAI recently announced their new consulting service that their minimum consulting client is a $10 million engagement and they're literally putting AI integration engineers to truly help people take it past and get it to the next level. What are your thoughts on that?

Jim Olsen: Uh well, I mean again it depends on what you need as a c a company because uh you know there's one thing about like running your own LLM or a customized LLM. I personally have a viewpoint I think we're going to shift a little bit where instead of rather than having everybody do uh large language models which are basically generalists for lack of a better term. You know they know a little bit about everything and there sound very convincing they know it all. Um and they don't necessarily do.

Um, you know, with the recent advances in model distillation techniques where basically they actually can use one of those generalist models augmented with very specific uh information to the task at hand and then instilling that knowledge into a small language model that you can actually run locally yourself that sets up the perfect kind of uh agentic architecture where now I have expert agents just like you don't have one worker who knows it all. I mean too many people think CTO's or you know you know everything. No, you can't. It's not just not possible.

instead you are that generalist who can then find the experts and work with the experts who can dive to that next level deep leaper deeper and then you know when we actually look at that I think that's where SLMs will start to come in where you have these teams of agents basically not that different than humans who have specialized knowledge work together and then are supervised you know there's a swarm method and a supervised method but the are supervised to actually then do this in a lot more cost-effective manner because now I can run these SLMs locally obviously open AI doesn't want that But it's like I know that's where I see some some very interesting things going going forward.

Steve Taplin: Very interesting. So Jim, you know, you mentioned CTO's can't know everything. It's, you know, it's it's probably one of the hardest jobs out there over the past couple years. It's been very volatile.

CTO's are blamed for everything. They not only have to lead the engineering teams, the technical side, but also need to justify their existence to non-technical executives and financial professionals who can't even spell AI or don't really know what it means. And you've had such a distinguished career. How did you get your current role at ModelOp's?

Jim Olsen: Well, what I came in is initially entered uh the company. They were more of a consulting organization but kind of had the area of focus that I was interested in around model governance uh but didn't really yet have a tangible instantiation of that. There were some proof of concept code and some items that they used in their consulting business as accelerators but um to put it bluntly they weren't enterprise-grade. So I was brought in initially to build a team and redesign the product uh and get get a first version out.

So I'm I'm a very rapid prototyper. So basically I was able to put together a whole platform that demonstrated the art of the possible and get it put in place. But then obviously it became clear to my superiors that it was better I was doing more of that side doing the architecture doing the actual development guiding engineers and staying on top of the industry than spending my time doing reviews and tracking Jira tickets and things like that.

Steve Taplin: So we kind of decided to split the roles and they moved me into the CTO position to take that over and then uh we brought in a VP of engineering who was more targeted with the actual mechanics of running the people side and the ticket side and things like that and I concentrate around the architecture education of the engineers uh you know and all the different aspects of that but more technical back up. How did you get in the door at the company? Was it a LinkedIn job at a recruiter? Did you know somebody?

Jim Olsen: It was kind of we had worked with them in the past uh basically so is through networking basically and they became aware that they wanted to jump to this next level. Uh I was in a small company that got acquired by a big company. Things were tumultuous in the large company. So it was an opportune time that I was looking for another opportunity.

So they approached me around coming in as the VP role along with my other former colleagues who came as VP of product. So it's kind of a ready built team.

Steve Taplin: Okay.

Jim Olsen: So they brought it brought us in together to basically kind of turn it into a product company rather than a consulting company.

Steve Taplin: Gotcha. All right. Well, tell us about some of the current projects you're working on.

Jim Olsen: Well, the most recent work we've been doing is obviously the agentic space is extremely interesting as I mentioned. So that is kind of the wild west. I mean we've already seen some uh things where you know the MCP protocol is relatively brand new. Uh it has some security challenges etc.

And we've already seen things where people are accidentally disclosing information across MCP tools, not really thinking about what it's doing because you're letting the LLM basically have access to the tool. Uh the tool, if it has access to data, who knows what it's accessing and when? And it goes back out to the to the LLM and now where's the LLM send it? Does it send it to another tool that publishes it to Slack?

Uh there's a fundamental lack of understanding and protection in those areas. So what we've been working on is governing initial demonstrations are around uh MCP tools themselves uh and the models and their token usage and interacting with those tools. So understanding what use cases in your organization are using which tools at what time and also being able to through a proxy put in place what we call protection models that scan basically what's going through that tool and can block it. So things like I can scan for PII, I can scan for injection prompts, I can scan for whatever so that I understand who's accessing what and when and get a dashboard view of these are all the use cases that have used this tool and how many times and here's the response times and here's the things and by the way the LLM that use case is using is using this many tokens costing us this much money.

Is this use case worth it? So really starting to be able to manage full agentic solutions from a governance perspective.

Steve Taplin: Okay. And that's what the ModelOp's platform does.

Jim Olsen: Well, it ma manages everything from an Excel spreadsheet to a machine learning model to generative AI to agentic solutions now as well as a whole in a consistent single pane of glass view because this these same problems exist for if I'm just using a traditional classifier or a logistic regression or I have an Excel spreadsheet that does forecasting for the business and it like a lot of banks will use that as their methodology of proving they're solvent. You need to manage that. You need to govern it. And to us, a model is a model is a model.

But we have enhanced techniques where it makes sense for the different kind of specific models. So it's not just this. This is another part of a bigger picture. We do the holistic governance of all decision-making assets like this, machine learning, generative AI, etc.

Steve Taplin: Excellent. So talk to us about your engineering team. How many engineers do you have? What's your work model?

I'm guessing you're remote as you're in a cabin in Colorado right now.

Jim Olsen: Yep. Yeah, we are remote first. All employees. We do have a central We have central locations in Salt Lake City and Chicago where we do kind of have uh clusters, but we absolutely have people who are fully remote way off of that on that.

That's uh that came about as we were always a little bit remote but when the pandemic hit and everything went remote we saw as an opportunity to attract the best talent from wherever and so that way we can bring in the talent where they sit and not only hit one talent pool. So that's an important aspect I think uh going forward that some of the RTO misses is that you do miss then you now just reduced your your pool of available engineers massively whereas now I can get the best of the best and we can offer them a an environment that works for their lifestyle and they can come and work with us on that. We set up in and I've used this in multiple different companies now. We go into what we call squads.

Um it's basically one leader engineer plus uh two to three more more junior engineers, not necessarily junior but not the senior lead. Um and they have areas of focus. That doesn't mean that doesn't change but in any near-term situation they kind of become experts in a space and you can even have you know some swap between teams to uh help distribute that knowledge. they work on those spaces in specific and provide us the and work together fairly autonomously much like I was talking about with agents.

You know, you want experts in an area, you want them to do their job, how they best do their job and then uh you know, report back up and just be held accountable for deliverables, you know, on that kind of things on that. So, that's kind of our approach on that. It's a small company, so we all pitch in wherever needed at whatever time. Obviously, that was a support issue.

Everybody cross-pollinates on that so everyone understands all areas of the projects and things like that. And right now we have about 10 what I call hands-on product engineers that are working. We're expanding on that. But we also have customer support engineers.

We have sales engineers etc. And that kind of things that not necessarily related for deliverables on that kind of things. personally in the CTO right now I don't have any direct reports kind of on purpose because I've been seeing that but that's something we're looking at expanding on too because I do a lot of prototyping work uh getting things out there floating ideas seeing what makes sense because a lot of times you don't know until you build a little something and so I work on those and then uh work on uh reviewing with the engineers and kind of passing off what makes sense or incorporating the ideas into what they're doing etc and I do just general architecture view because I've had so much experience experience on that. So I kind of act as overarching architecture or if it's a really nasty problem, I'll I'll jump in and said I'm hands-on still like I'll debug a customer problem just like anybody else.

I'm not above that by any means. Matter of fact, I like it because it keeps me real.

Steve Taplin: Nice. What AI assisted tools from an engineering perspective do you like?

Jim Olsen: Uh right now we're we're a Java shop primarily. We do Python and we do JavaScript as well and that kind of thing but we work in the enterprise and we integrate with things like Jira Service Now OAuth providers we do a lot of integrations which is unique about us as well. So Java gives us a good reliable platform for that also enterprises are comfortable bringing into house. So you know they're not we fly through every single architectural review we ever get because of that.

So because of that we like personally like IntelliJ they're really good. I played with Claude Code integration into IntelliJ and things like that, but frankly for the kinds of things we're doing, a lot of the prototyping works great with the AI solutions and we use the Junie integration that IntelliJ recently put out. It's cost effective. It's not the quickest, but it does tend to give very reasonable highquality suggestions and outputs.

Uh, and it is agentic. They offer two modes, a chat and an agentic, which I like because sometimes I just want to write a regex and I don't want to manually figure it out. They're great at that. Sometimes I want an agentic approach where it's scanning all the remote things and things like that as we've just seen like I mean cursor did some good things but burns tens of millions of tokens and I don't know if you followed the Reddit Reddit thread on cursor lately but it they just upset everyone because they just want jacked up their prices to cover the cost of what it was really doing and sure enough you know now it's unaffordable to use.

IntelliJ is a just it's more conservative in approach but it's also cost effective and works well. We've also uh played with uh the Junie integration into VS Code as well as Copilot. The new stuff, the old stuff me um on that. They've got a new beta out that is more agentic in nature and is better and looks promising.

I'm kind of conservative when it comes to some tool choices because those can be oneway doors. They can uh really trap you in quick. So, you know, you deal with something like IntelliJ or VS Code or etc. They're not going away tomorrow.

We've already seen some churn in the industry around some of these ones that everybody loved up front. They're gone. Now we got a new one and it looks like cursor is not doing well. So I tend to be a little conservative on those kinds of approaches.

Steve Taplin: Well, then when Windsurf was supposed to be acquired by OpenAI and that fell through.

Jim Olsen: Yep.

Steve Taplin: That was one of definitely uh what's great one week or one month or one quarter isn't necessarily uh rapidly as everything's changing. It's it's a moving target. Yeah. Yeah.

And I just kind of feel like these solutions they all use the same like Junie uses Claude Code underneath the covers or actually you can pick the model they're all going to converge on functionality I believe on that kind of thing. So the major players may very well end up just owning and winning the space. So as a CTO here halfway through 2025 what's the biggest challenges you're facing now?

Jim Olsen: Uh keeping up with what's going on and just like I mentioned trying to pick the right bets. You know, there's so much stuff that has popped up and disappeared this year in rapid succession. So, what are the likely especially in the agentic space, what is likely to really take hold? Like, uh, you know, MCP really seems to be an item that is taking hold and that's one of the reasons we're investing in that now cuz that seems like it's actually going to stick.

But like, which agentic platform is going to win? I don't know right now. You know, and uh, you go all in on one and you pick the wrong one. Well, there was a whole bunch of wasted time and effort and money and things like that.

So, it's figuring out that balance of staying ahead of the curve from a product perspective, but not so far out over our skis that we're we're running down trails we have no idea where they go. Um, you know, so how do we pick that balance of what seems to be like a likely good bet on a horse, so to speak? And that's very challenging right now because there is so much hype. There's you get you have to dig in.

you see the reality behind the hype and you're like, "Okay, there's actually nothing there. There's so much being claimed to be agentic now that's not agentic." Um, so how do you weed through all this and pick the right horses to place those bets on? And when it goes this fast, that's a challenge.

Steve Taplin: So Jim, what you've been doing this a little while, you've had some great successes. What's one brutal lesson you wish you would have learned earlier in your career?

Jim Olsen: Well, uh, as said, I did learn it somewhat early in my career, but the biggest lesson I learned was there's the whole technical side, and that's great, but the best technical solution does not win. It doesn't say it can't win, but just because you have the best technology, the perfect architecture, the greatest implementation in the world, and it checks all the boxes and hit all the hits all the buttons. That's a very small part of actually making a successful company. It's then also your messaging, your people, your sales, your interaction and there's all technical aspects to all of those like making the technology palatable to individuals communicating it to a way that they will understand etc on that.

So there's an awful lot more to being a CTO than just knowing the technology and picking the right solutions. Absolutely that's your your bread and butter. But if you don't manage the people and you don't and convince your own people for instance that this is why this is the right pick. Do you agree not to agree?

Getting them all on board in the right directions. Novell went through this. Uh it would it was all IPX/SPX based and you know that's why Eric Schmidt came in to put in the TCP/IP and literally everybody pushed back against it and Novell ultimately failed because they were late to the market and adopting TCP/IP. Not to say they didn't do it, but you know, it's like these kinds of decisions and management of not only the technology itself, but the adoption of the technology, explaining in the industry, etc., will make or break a company, and you need to be involved in that as a CTO.

Steve Taplin: Jim, well, thank you so much for getting uncensored with us today. Man, you have been involved in the cutting edge uh throughout the past three plus decades of technology. It's awesome to hear your perspective, your thoughts, how you attack solutions. Uh, great content you gave for the show.

In closing, any final advice for tech leaders you have?

Jim Olsen: Well, you got to always keep up and you got to cut through the hype. Um, you got to figure it out, especially in this market right now. There's one way doors and there's two-way doors. Basically, oneway doors, uh, you got to be really careful on.

That's where you got to spend most of the time in decisions. things like database structures or hanging yourself on one particular platform or those kinds of things. You got to be really careful and conscious of those. Two-way doors, you can run a little fast, you can play, you can run loose and cut yourself a little slack on those, realizing, okay, we made a bet there, but we can backpedal on that.

So, always examine, is this a one-way door, a two-way door, and react accordingly.

Steve Taplin: Great advice, Jim. Thank you so much. That's it for this episode of Software Leaders Uncensored. I'm Steve Taplin.

If you like the show, hit subscribe. If you didn't like it, well, tell your competitors about it then. They could probably use the help. Remember to fail hard, win big, and build what matters.

See you next time, everybody.

Narrator: That's a wrap. Another great episode of Software Leaders Uncensored. Thank you guys so much for listening. I really appreciate it.

It's a lot of fun doing this show. I love talking with great tech leaders. As a side note, if you're interested, check out my new book, Fail Hard, Win Big. This is the story of how I built 30 companies, failed on 20, and turned 10 into multi-million dollar wins.

The common thread with all of them was custom software. Also, if you need software development help out there, my company, Sonatafy Technology, we deliver world-class software development engineers out of Latin America. We have a solid track record of helping companies launch better products and faster. Thanks again for listening.

Appreciate it and appreciate you.