Technology Trends and Cultural Differences: IT Traveler with Jonathan Chang, Silverpond CEO from Melbourne

We continue our newly launched IT series with CEOs, founders, and CTOs of software development and testing agencies from across the globe. This time, we’ll talk about Silverpond, a digital agency based in Melbourne that specializes in developing AI systems. 

Silverpond is aimed at transforming the way people see machine learning. Let’s discover some interesting facts about the dynamic business landscape of AI in Australia from the CEO himself!

Meet Jonathan!

In this conversation, Jonathan Chang, CEO of Silverpond, together with Redwerk’s founder Konstantin Klyagin, discuss technology trends, cultural differences, and the challenges they face in managing and growing their agencies. Read on to get some valuable insights!

First of all, I want to introduce Jonathan, who’s the head of the Silverpond agency specializing in AI and headquartered in Melbourne, Australia. 

I wanted to start with my favorite tech joke about artificial intelligence: how do you distinguish between machine learning and artificial intelligence? Machine learning is when it’s written in Python, and AI – in PowerPoint.

Do you see a lot of that happening in Ukraine and Eastern Europe as well?

Yeah. Well, it’s penetrated the entire technology market, so we are not trying to be the vanguard of the technology and get into every emerging technology because it will just kill us. Basically, we would just spread the resources and never accomplish anything. Our approach is that if the technology becomes popular enough, we’re getting a project with it. If we have to deal with it, we learn it or hire people who are proficient in it.

We, of course, have already worked with machine learning like Azure for a customer in New York. They wanted to calculate the probability of problems with the loans that they give out. It’s an investment fund.

I think there’s still a lot of hype around AI. It’s interesting to see what will happen in the next three or four years as some of these companies will have to deliver against those promises. I wonder if they could.

I like the recent article about AI where a Google engineer discusses what intelligence is. What does artificial intelligence mean? Because we don’t have a definition for it. Intelligence is the ability to learn anything. And now we just have tools focused on solving a particular problem. It’s not intelligence as it should be.

Tell me about the company. Why did you start it? You are the founder, right?

Yeah. We’ve had an interesting journey. I originally started as a software development agency, probably similar to the one that you run. This is going back about 12-13 years. We started the journey doing some .NET, but moved reasonably quickly into doing Ruby on Rails, which we did for several years.

It was popular.

Yeah. And then we started looking at a bunch of functional programming technologies. We did some work in Closure, we did look into Haskell but it was very difficult to get Haskell to stick. Through that journey, we arrived at doing data science and machine learning. I think throughout that journey we’ve been just looking for a way for us to differentiate what we do out in the market.

Well, that’s a very common recommendation to agency owners that you’ve got to find a niche. And AI is a very interesting niche. By the way, I’m also very interested in this topic because there is a niche market versus all things to all people, right? And all things to all people doesn’t differentiate you. Every book on agency business tells you that. For now, my niche is SaaS, but it’s still very wide. 

They also say that you are not losing your old customers or word-of-mouth because they’re gonna be referring to you anyway. It’s only the marketing that you need to change. You just need to put your niche keywords on the main page of your website and direct all of your marketing into that niche. Did you see that in your experience?

So, this transition into machine learning has probably taken us about four or five years.

Did you deliberately close all your previous accounts not related to machine learning or?

I think we just went through a natural attrition over the previous accounts. We still have some customers that we work with from our software engineering days. But I think we were not not actively chasing the software engineering business with them. That’s primarily because we’ve been shifting the way people see machine learning.

Why did you start the agency in the first place?

That’s a really good question. I have built a CRM for car retailers or car dealerships. That was in partnership with two other guys. When they sold that software to a software distributor, I just started freelancing. For me, it was a realization that when I was working for other people, I didn’t have the control over the outcomes that I wanted.

The outcomes primarily being the rate of development and learning about the software engineering craft. There’s nothing better than having your own business. You’re ultimately responsible and you can never blame your boss because you are the one. I think this responsibility helps set me on a lot of personal development.

Well, that’s a very good reason. How do you cope with stress?

Meditation and exercise – I have done both in the past. I grew up a Buddhist. I think it’s learning to accept, which is probably one of the hardest things. The part of what drives us is just not willing to accept certain things. But I think that’s also where stress comes from. You are trying to realize that some things are truly out of your control.

Do you think it’s easier for a Buddhist to cope with the stress coming from running a business or is it the same for all people?

I think it’s the same for all people. It’s just, you know, setting goals, dreams, and ambitions versus reality. That gap is the source of a lot of stress.

So, when did you start the agency?

It was about 13 years ago.

Pretty much the same time as I did. How many people do you have at the moment?

About 16-17.

Do you think that size is optimal or are you looking to grow it?

We are looking to grow it. What I’ve learned from the last couple of years is that we grew pretty quickly and we had some challenges. Prior to growing, it was basically me and a group of consultants. As we started to add people, not everyone could report to me. 

What’s the size of your team?

Well, my company consists of 60 people.

So you would have had to have managers.

Yeah. For the first eight years, I was managing all the accounts. When I looked back at that – it was burning hell. Then, I finally started hiring project managers who are responsible for their accounts now. I should have done this earlier because it’s a big improvement.

Yeah. I went through a similar phase. We’ve only put in a senior management team the last couple of years. It’s been going through the transition and the cultural change required for us. That was an interesting thing because we’re pretty much a tight-knit group of colleagues.

In my organization people come to me whenever there’s a non-standard situation or a conflict to resolve. I’m kind of a repeater in these situations. I also try to document processes as much as we can so that it’s easier for new people to get into the company culture and its practices. Plus, it’s easier for everyone to know what to do.

Have you ever struggled with that cultural shift where people had a personal relationship with you?

Of course. There’s stuff turnover and everything. We had to lose a few people, which was not easy.

We’ve been through the same thing as well.

This is quite normal in a transformation like that. Being a new company, who was your first hire? Or were you just a group of consultants working together? When you were a freelancer, how did this second person start working with you?

I did hire one or two people, but probably the key person that we added to the team was a guy called Andy. He worked with us for quite a while and helped us transition into a machine learning company as well. He’s now off doing his own kind of crazy startup – building the Matrix. 

So, the idea that they’re currently working on computation on living neurons. Instead of imagining a silicon ship, you have real biological neurons. And then you can program them by constructing simulated environments.

That’s very cool. Where does he get neurons from?

That’s a good question. I don’t really know. They do have a bio wet lab and they’re putting all the infrastructure in place.

I’ll ask you for a link to that.

I’m more than happy to connect you with Hans – CEO, and Andy – CTO.

Sounds a bit crazy, but there should also be some science behind it.

There is definitely science behind it. No doubt. What he wanted to try to do with his career, I couldn’t really give him that. Even though we had transitioned into a machine learning company.

That’s biotechnology, a completely different domain.

Yeah. And it’s a very speculative kind of idea. Is there a market for people who would want to do computation in that particular manner? But you’re young and you should definitely go and chase after some crazy ideas, right?

That’s very cool. At least it would make good marketing and news.

Absolutely, it does.

How do you structure your company now? What’s your structure?

We have four different areas: business commercial, operations, engineering team, and science team. Probably, what’s not usual in agencies is this kind of science and engineering split. But we did that deliberately since the science team has a very different focus to what the engineering team does.

Those heads of departments, are they hands-on client projects as well? Or do they only do management?

So, given our size, we are all pretty hands-on in terms of interacting with customers, ensuring project delivery, etc.

Do you have any products or do you work only like an agency providing a service to your customers?

We have come from that agency and services background and part of this transition the last few years has been into being a product company. We have a product called Highlighter. The product helps with the development of machine learning projects and focuses on the development of the training and evaluation data sets. We don’t have technology that is like a TensorFlow or PyTorch. Instead, what it’s focused on is the management of the data that allows people to train these models.

It’s our belief that this data is quite important, particularly since a lot of our customers work in regulated fields. When they come and try to understand how these models work, inevitably they need to come back to the data used to train that. So, Highlighter is focused on those particular sets of problems.

It mostly generates data sets, right?

Yes. So, customers would be collecting data in some form or another. But when we started our journey five years ago, a lot of the way we built these machine learning models was on a data scientist laptop. 

They run up their IPython notebook and then how do you manage all this data? How do you manage all the different stakeholders? If we needed doctors to provide annotations and review of some of that training data, how do you enable that when it might be seen in one person’s laptop? Some of the infrastructure needed to be in place to help manage and generate this training data. That’s what that tool does.

But that is something like using your services yourself?

It’s actually both now. We do have customers that use it without our services. Some customers engage us purely on our services and some customers combine that.

Is the product a gateway to your services? Because it’s a very common model. You have a product that people start to use and then to customize it or to use more efficiently, they have to hire you to provide professional services based on the product. Is that something that you do?

We have been asked to provide professional services around our products, whether a product is the gateway for services or services are the gateway for our product. I don’t know which way is the correct lens. And I think we’re still discovering as we go right now.

Speaking of growing the team, is it difficult to find an engineer, data scientist, or any other relevant candidate in Melbourne?

The Melbourne market is reasonably competitive and I think there’s been an entrant of a number of well-backed US and other companies into Australia. They’ve been looking for talent.

How long does it take you to find an engineer or a software developer?

2-3 months.

That’s tough. How did you personally learn to delegate in your company?

Delegating is such a challenging conversation. I think the first step is to work out what people are capable of taking on. One of the biggest challenges or mistakes that I’ve been making is not having a clear idea of that. If you delegate them too much and it’s too hard for them, then you basically set them up to fail. If you delegate too little, then they don’t get any value of why they’ve joined your team.

Well, I have an insight about the CEOs’ and managers’ duties. First, I try to understand the process myself. And when I have an understanding of what steps need to be taken to get this result and how this particular role works, then I can hire a person who’s more experienced than me and who’s much better at this activity so that I could delegate the duty to them.

But how do you assess that a person is actually better than you?

Well, there is a matter of trust. Also, there are some HR ways to assess that. There’s a resume and you can ask the previous employer for recommendations.

A common gap for technical people in particular is when they go from being an engineer, for example, into management. How do you work out? How much to delegate? How much training they might need to actually be successful in this new role?

I think it must be related to their previous experience. It’s also better for them to keep developing in this particular direction and not trying to change directions from purely technical roles to purely managerial roles, because you’re making a bad manager out of a good software developer.

Do you set up a completely separate track for developers in your team that might go: “You know what, I’m just going to be a very good developer who gets paid very well and not worry about my career growth as part of being in management.”

Well, many developers prefer to stay developers because they’re better with computers than people. For me, people were the most challenging thing because my background is computers. I’ve been programming since I was 18 years old. 

What we practice in our company is a horizontal structure where we don’t have heads of departments or managers. We have account managers/project managers who are also a part of a flat structure. So, if you ask someone who’s your manager on this project, they’ll say that on this project this guy and on that project that guy.

Do you have someone who is responsible for career development for an employee who does well?

Yes. That’s a blend of the HR and project management team. Whenever someone wants or is required to improve their skills, they set up an educational plan and this person has to go through it to acquire the skills necessary for the upcoming projects.

How big is your HR team?

It’s just two people: an HR and a recruiter.

Since it’s an agency, people are probably the most important thing. I guess a recruiter can help a lot with that.

The role of a recruiter is very important. I’m super happy with my current recruiter. Once I heard how people describe the difference between the HR and the recruiter roles. A recruiter gives birth to babies and HR raises them. The recruiters are responsible for the new hires. The HR role is to keep them happy and facilitate all kinds of development, activities, and loyalty in the company.

That’s interesting because I think we don’t have a dedicated HR or recruiter. We tend to rely on the senior management team to perform the recruitment function.

Of course we rely on our senior developers and project managers to interview, hire, and get people interested in working with us. But, from my experience, you can’t do that efficiently without building an HR ground, without maintaining a certain company image on the job market.

But I see that you’re doing some activity in this direction as well like those meetups and conferences. That’s a great way to attract new candidates.

Is it something that is happening a lot in Ukraine as well?

We were among the first ones to set up these kinds of formats in one of our locations. We have two offices – in the capital of Ukraine, Kyiv and in Zaporizhzhia, which is 500 kilometers south from Kyiv. Zaporizhzhia is smaller than Kyiv but it has two technical universities. 

I think we were the first to start doing meetups. Now, it’s picked up by other companies. We call it developer meetups. We invite people to our office to deliver speeches and presentations. It’s quite fun.

Do they have specific technologies for each of these?

Normally it depends on the presentation that we have around web development, backends, etc. So, we announce those speakers and presentations and based on that people decide whether it’s interesting for them or not.

Who looks after those meetups?

HR and office management. They prepare everything. They have a checklist for every meetup: what needs to be printed, what needs to be arranged, how the room should look like, attendance register, how we accept the registrations, and who meets them at the entrance.

What kind of technology trends are you seeing up in Ukraine?

We work on the global market. We don’t work on the Ukrainian market because it’s very small and not mature for services like that. Sometimes we’re getting some advice from local companies. So, normally our customers are either from the United States or Western Europe. But recently we have seen interest in our services from very unexpected parts of the globe. 

We have a few customers in Africa. We can work with them because our entire process is set up in such a way that you can work with remote customers anywhere. Of course, we also invite them to visit us and sit with us in the office. But mostly the work is done remotely.

So, the trends. Well, everything’s in the cloud now. Compared to when we started, it was all self-hosted. Also, Python, .NET are coming back.

That’s true.

There will be a lot of work at the beginning of this year. Unexpectedly, a lot of new customers, which is good.

Do you think that the American and European markets are doing more outsourcing now? Are they trying to bring it back in-house?

I think that everyone is comfortable with the idea of outsourcing because it was more difficult to persuade people to do outsourcing in the past. It’s very common now. Everyone has done it. The question is the choice of a vendor.

Yeah. In the Australian market we go through the cycles of where these larger businesses try to outsource everything and work with lots of different vendors, and then at some point they go like: “Oh, we should really insource everything”.

It’s always like this with big enterprises and corporations. It goes up and down: outsource everything, take everything in-house, outsource, and again in-house. That’s normal for any big organization because the management changes. For example, there’s a guy in charge who’s a great believer in outsourcing and then the next guy says “No, I want to see them all in the office”. 

Speaking of customers, are all of your customers Australian?

Most of them.

If not Australian, from where?

Singapore and the US.

How is the Singapore market?

It’s a big financial hub. The people that we work with have a series of startup interests. They are basically engaging us to help them with some of their startup projects.

All right. How did you get this customer?

They found us through our website. We largely focus on the Australian market. We feel like it’d be quite challenging for us to enter into a Singapore or overseas market.

But now that you have a customer there, if you make them happy, they would recommend you further. You can expect more leads. This is how it always happens.

That is absolutely true. We can’t imagine Australian service wages as being competitive in the global market, compared to Ukraine. Our differentiation will have to be something else. Probably primarily through our product.

So, how do you normally acquire your Australian customers?

Through word-of-mouth. We’ve had some success in our efforts through the meetups and conferences.

Do you have a sales team?

So, that would be Ed and myself. Ed will look after the commercials.

An insight from me. I saw a big growth in the amount of customers when I hired a salesperson. I had to change a few salespeople and I had to reformulate the entire sales department. But I’m super happy with the current setup. Today, I got a report from them on what they have done in the previous month – 2 new customers, 2 new countries.

That’s fantastic.

It came just out of their own efforts and out of our inbound marketing where we invest into marketing and search engine optimization. This is one channel through which people find us. Another channel is active sales – cold messages. It’s not very efficient, but you have to do that. Otherwise, you depend too much on the word-of-mouth and whether your customers are willing to recommend you further or not.

I think that’s something that needs to be independent. First, it may be an investment, but then it pays off because it gives more customers.

So, I guess right now it’s me and Ed. He drives a lot of effort into the sales of our business. Where are you planning to take your career? Now that you’ve started a successful 50-60 person agency, what’s the next phase of your career?

That’s a good question. At the moment, I just enjoy doing what I do. It’s been very interesting and good for my personal development because I get to learn a lot. I have to switch hats all the time. On the same day, I may be doing marketing, sales, project management, and some technical details of this or that project in completely different industries.

 I think I still like it. My ambition would probably lie in the product development area, even though I have done this in the past. When I was 15 years old, I was developing software. Do you remember the bulletin board systems?

Yes.

One of the softwares that was used to run a BBS was developed by me. Back in 1996, it was very popular, at least in the post-Soviet states. Right in Eastern Europe, I had a big user base. Every evening I was sitting at my computer, processing feature requests from my users, and working on my own improvements. It was a very popular piece of software. The only competition was RemoteXs and Maximus.

My software was called Tornado BBS. It was purely product work.

Are you interested in probably doing that kind of thing?

Yeah, probably. We are trying some ideas internally. We’re trying to work on smaller products. I know it’s not very efficient in the agency model because once a customer wants more resources, they are just snapped from those internal projects and never put back. Recently, we’ve hired a project manager with the product owner background who would focus on those. I hope we can set up a process for that as well.

Now I know that it’s been challenging for us to make that transition ourselves.

I’m also looking to draw certain verticals and maybe convert them into separate brands. I have a very successful example of that – my QA departments are a separate brand. Redwerk does full-cycle software development for SaaS systems and QAwerk – manual and automated quality assurance.

I want to add penetration testing or maybe security testing. Then the same way we could package in the market our UI/UX design team, which is anyway there. Also, search engine optimization. You can do a lot. 

How do you feel when people leave the company?

When I first started, it was very hard. It’s almost like a part of the family. I think recently I’ve taken the view that it’s about trying to find the best fit for everyone. For a few people that have left, I’ve helped to find jobs in other agencies or companies.

I was the same. I was taking it very personally. Now, I came up with a few reasons why it is even good for the company. First of all, it’s an opportunity to document a certain process for a role, and then you get someone sooner or later who can bring you new ideas and shake it up a little.

I agree. I think one of the things I do look at is how to inject a sense of ideas and energy. Probably the best way to do that is to bring new people in.

 Yeah. There are burnouts and people get bored, no matter how interesting that was at first. And you can easily explain it to customers because everyone has staff turnover. If you lose a person on the project for some time, it happens. You can’t prevent it. There’s no cure from that.

How do you start a project? Let’s say there’s a new customer, what are the steps to start the project?

We do a bit of discovery with the customer. That’s primarily because machine learning is fairly new and a lot of our customers don’t even know where the value is. Getting that value conversation is pretty important for us to continue doing machine learning with these customers. 

That initial discovery phase sets up for whatever proposal that we end up putting in and the setup required to get a project up and running.

Then you assess the project or how do you also communicate the budget required for the project?

We have taken two different approaches depending on the sophistication of the customer. The sum is estimated from the effort and cost-based perspective. With some other customers, it’s done more from a value-based perspective where the customer wants us to take on risk. It becomes packaged like a fixed-price project.

You still do that? Because we dropped that about 7-8 years ago. It was burning hell. Now, we give a rough estimate, and if something unexpected pops up or if there is a new feature request, our project managers are trained to communicate that to the customer. If you want this new feature, it’s going to increase the budget.

From the software engineering side, it’s well accepted in terms of how that works. I think a lot of customers, at least in the Australian market, are still experimenting with machine learning. They want to understand some tight bounds around its delivery. We just take on some of that risk to try to help this particular market.

So, we estimate the amount of man hours multiplied by the hourly rate.

We did software engineering very much the same way.

Well, I’m also very interested in cultural differences. You’re saying that most of your customers are from Australia, Singapore, and the USA. How do you see the difference between those three locations?

I would say that there’s a different kind of focus between working with Australian and overseas customers. Some of our overseas customers seem to be more cost-focused, whereas some of the Australian customers are more risk-focused. They may be more concerned about some risk exposures that might be happening on a particular project. I don’t know whether that’s because Australian customers tend to be more enterprise customers, whereas our overseas customers tend to be more startups.

More open to ideas.

Yeah.

Is there any risk mitigation that you practice? Any particular approaches to that?

In machine learning, the challenges are the things that are outside of your control. For example, when you train the model on data that is collected, typically we might not control how that data is collected.

And if it’s wrong, you get wrong results.

Yeah. We don’t control exactly how it gets annotated. We may not have that kind of exposure. Often customers want to know something like “Hey, can I get 99%?”. It’s very difficult to promise that kind of thing. There’s a lot of expectation and risk management around various parts of this process that we have to put in place.

Probably a lot of explanation is needed to explain that this kind of a solution is not gonna give you 100% accuracy.

Yeah. We have to set those expectations. Particularly, during the proposal phase.

When was the last time you introduced a new practice in your company like a new technology or new vertical?

We’ve transitioned through lots of different companies. I think I’ve mentioned Ruby on Rails, functional programming, and then my data science machine learning. I think the hardest thing has been to bring the team along to some of those things. I’m curious to know like how you guys would have had to grow these kind of different capabilities, assuming you added .NET, PHP, and Python over time.

Well, we started as a Java shop doing e-government solutions for the company in the Netherlands which was a governance vendor for the entire country. We helped them develop products. Then we had to add new practices, so we just hired people with experience in those technologies. 

But we also have a few champions who let’s say when they finish their last project they’re like “I was doing .NET, Ruby. What do they have here? I wanna learn that!”. They would just jump on new technologies. They’re very eager to learn that. But that’s a minority. The rest want to develop in this particular direction. They say “I want to become a PHP guru, knowing all the frameworks in PHP. That’s my career path that I’m seeing.”

I’m fine with that. But if they have certain gaps in particular projects, that may be a problem. Of course, we’ll lose this developer.

But do you ever find that you need multiple streams of business to actually feed these capabilities? For example, if you win a large Java project, you need more Java people, but then you don’t have any work for your PHP people.

Exactly. So, we are trying to manage that with our marketing and sales activities. Whenever we have a gap or less projects than resources on a certain technology stack, we try to intensify our efforts to sell that particular expertise. It doesn’t always work out. It’s not very predictable. It’s not a very stable outcome, but you have to try that. Also, if you get to choose the technology for a new project, then it becomes easier because you have enough resources.  

You’ve chosen technologies that all have pretty big use bases.

Yeah. Allright. Have you heard anything about Ukraine?

What about colleagues from Ukraine?

What do you think about them?

What do I think about him or Ukraine?

Him, Ukraine, and the Ukrainian developers. What do Australians normally know? That’s what I’m interested in.

I don’t think a lot of Australians probably have exposure to Ukraine. They probably think it’s just a part of Soviet culture. Ukrainians are typically perceived as good engineers and mathematicians. I’d like to visit because my colleague is from Ukraine.

Please come visit us.

My wife is French, so I do have a reason to go to Europe quite a bit. Maybe I’ll drop by.

When was the last time you went to Europe?

In April, last year.

Because it’s quite a trip from here.

It is a trip. 38-40 hours to France – quite long.

But it’s Australia. It’s far away from everything.

Yeah.

So, we welcome everyone to visit Ukraine, visit us and our office, go out with us to experience the nightlife, the bar scene, and everything. Next time you’re in France, please do take some time and come to Kyiv, at least for a weekend.

Absolutely.

All right. Thank you for the conversation.

Likewise.

Work With Ukraine

The interview was recorded before russia initiated the largest armed conflict in Europe on February 24, 2022. Despite the war still raging on, Ukrainians persevere in a new reality.

Although russian terrorist forces were trying hard to destabilize the country’s economy, the Ukrainian tech sector keeps delivering through wartime. We stay operational and are eager to develop new partnerships. Redwerk would deeply appreciate continued solidarity and support from international clients.

By choosing Ukrainian IT service vendors, you can strengthen ongoing support and help keep our country’s tech sector afloat during these challenging times. And we in turn will ensure high quality of service. 

Stand with Ukraine by working with Ukraine!

Looking to support IT vendor from Ukraine?

Learn how Redwerk manages war-related risks.
Please enter your business email