DataTalks.Club

Freelancing in Machine Learning

Season 4, episode 8 of the DataTalks.Club podcast with Mikio Braun

Books:

  • Million Dollar Consulting by Alan Weiss
  • Built to Sell by John Warrillow

Links:

Did you like this episode? Check other episodes of the podcast, and register for new events.

Transcript

Mikio’s background

Alexey: This week, we'll talk about freelancing and consulting. And we have a special guest today — Mikio. Actually, Mikio was my teacher six years ago. I was studying at TU Berlin for my master’s and Mikio was one of the teachers of the machine learning class. I was very lucky, because, after that year, he left academia and joined the industry. So I was able to witness his teaching. After working at TU Berlin, Mikio joined Zalando. He worked there in many different roles. He spent quite a few years there. Right? (1:59)

Mikio: Four and a half years. (2:54)

Alexey: And then he joined GetYourGuide. He worked for some time there. And right now, Mikio is working as a consultant. He consults companies on how to make the best use of machine learning, looking on the whole picture from technical infrastructure to product development. Welcome, Mikio! (2:56)

Mikio: Hello! Thanks for inviting me and for the kind words about my teaching. (3:14)

Alexey: It was really great. I remember it was the sixth or seventh lecture of machine learning. Professor Müller was on a business trip and you were a replacement. You came into the class, you looked at a topic, and said, "Linear regression on the sixth lesson? Seriously?" (3:21)

Mikio: Yea, I remember that. We also did a practical course on big data. (3:40)

Alexey: Yeah, that was the second semester, the winter semester. That was fun. I did talk a bit about your career already, but maybe before we start with our main topic, you can also tell us about your career journey so far. (3:48)

Mikio: Originally I wanted to become a professor. I was a researcher. But then the only goal you have is to become a professor. I was writing papers, supervising PhD students, doing conferences. I went to NeurIPS five or six times. Originally I studied computer science. I always wanted to build something and at some point, writing these papers became less and less fulfilling. It's also really hard. For these conferences, the acceptance rate is less than 20%. So you do a lot of work that is never published. (4:10)

Alexey: So this bump boost [an algorithm that Mikio invented], was it ever accepted? (4:55)

Mikio: I don't think so. (4:58)

Alexey: How many times did you try to submit it? (5:00)

Mikio: Like two or three times. It was a nice algorithm. At least, there's one guy who wrote his master thesis trying to parallelize it on a GPU. So at least it was good for something. Then I had this streamdrill / twimpack project, a mini-startup on the side, which we did it for a few years. We did tech pilots with some companies. So it exposed me to this world. The funny thing is, if you're inside the university, everybody tells you "this is the best place to be". And then you start working with other people, and you think "this is also interesting". Then I went to Zalando, and there I really learned a lot. I continued trying different things. I started as an engineering manager. Then they had this principal role, so I thought, "I want to be more technical", and so on. I look for opportunities to learn something all the time. (5:04)

What Mikio helps with

Alexey: Eventually, you ended up being a sole consultant. What do you help with? What do you consult about? (6:13)

Mikio: I help companies to do machine learning in practice. I worked on so many different aspects: I created algorithms as a researcher, I organized teams, I worked on a machine learning platform team at Zalando as a consulting architect. This principal role was a bit like internal consulting. I always start with a technical problem, but then I wonder, "what is the real problem?", and then you end up thinking about organization and things like that. So I help companies become better at that — whatever it is. That was the original idea. (6:23)

Alexey: So you're like a principal data science / machine learning person to hire? (7:18)

Mikio: Yes. One decision I made early on is that I wouldn't be hands-on myself. I think it's a very different kind of work. If you're doing data analysis yourself then you'll have to put on a lot of hours. So I'm more on the mentoring side, so I can also do a couple of projects in parallel. (7:24)

Moving from a full-time job to freelancing

Alexey: How does it work out for you? (7:53)

Mikio: So far, it's good. When I decided I wanted to try this, I quit my job at GetYourGuide. I thought that I first need to write a lot of blog posts for 2-3 months for promoting myself. I thought it would take some time. But actually, I immediately got my first project. And then another one. So I stopped working, then there was a weekend, and then I started a consulting gig. And there was no time between. It was a bit surprising. (7:55)

Alexey: Did you at least write something on LinkedIn, and this is how you found a client? (8:47)

Mikio: Not till January. I started in November. When I started, people I knew from Zalando times found out that I quit. They asked, "What are you doing now? Why don't you join our team?" I said, "No, I want to try consulting". And then a few of them said, "If you want to do consulting, we can try that". That's how I got my first gig. In retrospect, it's not that surprising. I knew a lot of people. If you're at a company like Zalando, you get to know a lot of people. And then they move on. So you have a good network already. And now everybody is looking for data scientists, they are very hard to find. So, it's not as hard as you think it is. (8:54)

Finding clients and importance of a strong network

Alexey: So it's not as hard as you think if you have a strong network. (10:08)

Mikio: Yes. I think it's the key. Also, you work with many people. If a project works well, then somebody recommends you. It's not like you build a great website, and then you sell it, and then people come. Consulting is a lot about trust. It's always better if somebody said that "I did a project with them, it was great". It's not uncommon that it works like this. (10:12)

Alexey: So this is how it happened to you. You found the first client through your network of people who worked with you at Zalando. You found them — or they found you. Then you worked for them, and they recommended you to others. Is this how it started? (10:53)

Mikio: Yes. They didn't recommend me. But there were a few of these things and they became the first few projects. In January I posted on LinkedIn that I'm doing this now. This career change on LinkedIn always get a lot of engagement. There's this thing that says "I'm working at a new company". And everybody says, "Yeah, Great! All the best!" On LinkedIn, liking is like retweeting. It's shared quite widely. (11:12)

Alexey: Also, if you update your profile, everyone in your network gets a notification. (11:48)

Mikio: That led to a few talks, but no new project at that time. But then I started doing lots of talks. I gave a talk here, I did a couple of podcasts. Through that, this summer, I got a large number — an okay number — of new inquiries. (11:55)

Alexey: By "okay" you mean more than you can handle? (12:28)

Mikio: Yeah. Workload-wise, it was okay. But currently, I have to think about what kind of projects I want to do. (12:31)

Alexey: So now you have projects to choose from? (12:46)

Mikio: Yes. In freelancing, you actually have this freedom. You can say "I won't do this". And if something's not working, or you're struggling, you can think about it and try a different kind of project. This was also pretty constant. In the beginning, I was like, "this will take a few months until I have income again". Then suddenly, it's okay. But what if it stops? And then, at some point, you have too many projects. (12:49)

Alexey: So, more things to worry about compared to a usual full-time job? (13:34)

Mikio: Yeah. (13:40)

Alexey: How do you usually find clients? Or they find you? (13:41)

Mikio: Yeah, they find me, so far. As I said, I did a bit of marketing here. But then the question is, "what do you do then?" You have a first meeting to understand what they need. (13:47)

Alexey: So to find clients, you need to have a good network. Then you need to give a couple of talks, so people outside of your immediate network know you exist. And that's enough to actually get started. And then the word of mouth helps to get new clients. (14:16)

Mikio: It's actually easy to start up. If you have one or two projects, then you’re already making enough money. You don't have to bootstrap for two years and not make any money. Transitioning was smoother than I thought. Then you see how it goes. Also, if a client is good and everything works well, they can extend the project. (14:47)

Building a network

Alexey: I want to ask you a few things. Let's first start with the network thing. You said starting wasn't too difficult for you because you had a good network. By the way, there is a message from Felix Müller, saying, "Greetings from one of those ex-Zalando colleagues". (15:28)

Alexey: For people who want to start freelancing, how do they build such a network? If they want to start freelancing in a couple of years, how should they approach building a network? (15:49)

Mikio: First, you can start in the company where you are. It's quite natural to me. (16:03)

Alexey: Hang out by the water cooler? (16:20)

Mikio: Yes, hang out by the water cooler. I sometimes meet for coffee. Not just in the team, but maybe there's somebody from another project. You say, "maybe we can have lunch together" or coffee, just to get to know each other. Also, I always ask people whether they know somebody and then they introduce you to them. You can have this typical one-hour networking meeting. Let's talk a bit about that. (16:21)

Mikio: You have to find people. You can ask people in your network whether they know somebody interesting. If you see an interesting post on LinkedIn or Twitter, you can connect with people and say, "This is interesting. Maybe you can have a meeting?". I'm also getting requests like this. There's also an app, which is called "lunch club". You register there and say what you're interested in. Then each week, you get matched with somebody. You can also go to meetups, and just talk to people. (17:04)

Mikio: First, you get the contact. Then you meet for one hour. And to me, it seems there's always a structure. Everybody does an intro about themselves, you tell a little bit about the history of what you're interested in. That takes like half an hour. Usually, in that time, if you see some common things, you go deeper into these topics. And then you're done. If you do this for a while, you get to know a lot of people. on. (18:08)

Alexey: So you said lunch club, LinkedIn, and meetups. I guess now it's a bit more difficult with meetups. (18:54)

Mikio: I think it's totally okay if you see a post on LinkedIn, just make a connection request. Then just say, "this looks interesting, if you want, let's meet". Just ask them. They can only say "no" or not get back to you. But people are quite a bit open to this. Maybe more open than you think. (19:02)

Initial meetings with clients

Alexey: The other thing you said, "Now you get a client. The interesting part is what happens next". First, you have this intro call. You figure out if you can help the client. Right? How does it look like? Let's say you get a LinkedIn message saying, "Hey, Mikio, I've heard you're freelancing. Can you help us?". What happens after that? (19:09)

Mikio: A few things. Usually, there's one meeting and then sometimes they say "You also need to meet these people before we make the decision". Sometimes it's quite a few meetings for which you don't get any money. (19:52)

Alexey: You don't bill that? (20:09)

Mikio: No. You try to build some trust, to see whether it actually makes sense. If you get a client, and this would be a project that takes a few weeks or months — usually for quite a lot of money. So this is the kind of unpaid marketing that you have to do. When I started, I had these mentoring projects — working for a few days, more like a few meetings here and there. Over time I realised that all this other work is also work. You don't get money for it. (20:15)

Understanding what clients need

Mikio: We also need to understand what they want. And it's also important to understand what their problem actually is. (21:37)

Alexey: What they want and what they need. (21:47)

Mikio: Yes. They are already looking for somebody who does this and that, but that's already the solution part. And then the question is, "Why do they want that?". (21:49)

Alexey: Like, "We need an expert in deep learning". But why? Then you start digging. And then turns out that maybe a logistic regression would solve their problem. (21:59)

Mikio: Lecture number six, right? [laughs] (22:12)

Template for the offer (Million dollar consulting)

Mikio: There's one book with a very funny title that somebody recommended. It's called "million dollar consulting" by Alan Weiss. That's a good book. I'm not sure whether you can take this and make one million dollars. (22:18)

Alexey: I was going to order that. Now I'm not sure. (22:37)

Mikio: He gives a template for the offer, which I found quite good. First, you take time to write down a summary of what you understood. What is the problem they want to solve? What are you offering? Then you can also talk about your fees. Every time I do this, it's actually quite insightful. It's also good for them to check whether you really have the same understanding. Sometimes they say "When can you start?" and it never gets really explicitly written down. (22:45)

Alexey: So the outcome of the initial meetings is to understand if you can help them and what they actually need. Then you write a summary and decide on your hourly rate. Then you send them this email, and then they either agree or not. Right? (23:20)

Mikio: Yes. You can also do it in a Google doc and have them comment on it, and discuss it a little bit. (23:40)

Deciding on rate type: hourly, daily, per project

Alexey: We already have a question here. How to decide on this rate? The question is about the daily rate. We had some prior chats about this, you also mentioned something about a trade-off between pay per day or per hour or per project. Maybe you can also talk about that a bit. (23:52)

Mikio: First, I should say that I'm really bad at negotiating. If you're a future client, you should know. The natural thing is to start with an hourly rate. It's very transparent. You have Google Sheets where you put in all the hours. Then, at the end of the month, you write an invoice. But as a downside, this is a very strange incentive for you. It would be great if you get more money if you help the client, and not if you just work more hours. Sometimes it doesn't make any sense. I also don't want to work more hours. And then you suddenly realise you're actually not working as many hours as you thought you would. For whatever reason — because it's easier, or the project is moving faster than you thought, or people don't need that much help. (24:15)

Mikio: But if you go for the hourly rate, how do we can come up with this rate? In the "million dollar consulting" book, he proposes to talk about the value of the work you're doing. Then you say, "I go in, I do this logistic regression model, you save 100,000 euro. So let's say 10% of that" or "5% of that". This is a good argument, but sometimes it's hard to figure out the value of the project. Also, many people think in terms of salaries. (25:19)

Alexey: It's also risky. How do you know that your project will succeed? You're betting on the success of the project, but in machine learning, it's not always the case. (26:05)

Mikio: That's true. The book says you ask the client "How much money will this make?". They say, "Could be between 1-2 millions". Then you take the lower bound. So you say "You think it'll make at least 1 million?". I've tried. I know people who managed to do that. But I think it's a much tougher discussion. Also, if you do this, you can never talk about hourly rates. If you talk about hourly rates, they will always try to say "Okay, if this is 10,000, how many hours is this work?". And then you can't get away from it. (26:17)

Mikio: And then there's another weird thing. You ask for 10,000 euro over three months, so each month you get a third of it. Or you ask half first and then half again. But with this fixed-rate, some companies are suddenly afraid that this might look like employment. If you're paying somebody as a freelancer, but it's like employment, then as a company, you have to pay health insurance and stuff like that. (27:04)

Alexey: They are afraid that the finance authority will come and say, "looks like you're evading taxes". (27:51)

Mikio: You have this guy, you're paying him a monthly fee, which looks like salary. So now you have to pay all this health insurance. So this is the downside. (28:07)

Alexey: If somebody wants to start freelancing, but they have no idea what the market rate is and they don't know how much other freelancers make. Then it's difficult to come up with this hourly price. How to come up with it? Do you ask the client, "How much do you want to pay me?" When you're starting, it's difficult to come up with a number. Do you have any tips? (28:26)

Mikio: I asked other consultants when they make. I was lucky because my first client was somebody I worked with previously. At some point, he suggested that I increase my rate a little bit. If you're hands-on and working a lot of hours you make less per hour. It's linked to the impact you're making. But then you also have to see whether it feels right for you. (28:53)

Alexey: So you don't have a daily rate for your projects now. Or how do you do this now? Hourly rate, daily rate or project rate? Or it really depends? (29:37)

Mikio: It depends on the client. For older ones, I have an hourly rate. You can set your prices and if you can agree on something, it's good. The advantage of a fixed rate is that you can focus on doing the best work. Sometimes they're like, "He's so expensive, let's try to use him less". So you can just focus on work. And sometimes it's more work, sometimes it's less, but it's also less of a hassle. Also, if it's an hourly rate, usually clients ask for the weekly estimate. Because they need to do a budget for that. So actually the fix rate doesn't really hurt them, because then it's already budgeted, and they can just spend it. (29:51)

Mikio: But at the end, you have to ask yourself, "How many hours do I want to work?", "How much money do I want to make?", "When does it feel okay or not okay?". Also, when you come out of a full-time job, you know what you would earn as a full-time employee. (30:02)

Alexey: This is how much you aim to make, at the very least. (31:29)

Mikio: It should be more because you always have to pay more for your health insurance — you have to cover all of it. And you have this risk that there is no project. (31:34)

Alexey: So you need to think about the average, the expected value, right? (31:48)

Taking vacations (and paying twice for them)

Mikio: Yeah. And your vacations, right? Nobody pays for that. (31:52)

Alexey: About vacations. I know you recently came back from vacation. I imagine that it's very difficult to take a vacation. I work as a full-time employee, I know that I have 28 days of vacation. I can just say, "I'm taking 10 days or 20 days. See you there!". And I get paid for this. This is a paid vacation. For freelancers, it's not the case, right? (31:59)

Mikio: Yes. You don't work, you don't make money. (32:25)

Alexey: So psychologically it's very difficult to pack your things and go away for a month. (32:29)

Mikio: Yeah, exactly. (32:35)

Alexey: How do you deal with that? (32:37)

Mikio: I thought about this. Actually, I'm paying double for this, because I'm paying for the vacation, and I'm not making money. (32:40)

Alexey: If you're working from vacation, then you don't enjoy the vacation. (32:50)

Mikio: Some people go on vacation and have just this one client. I have a few. If there's not a lot of meetings, you can still have them. But that's the point. You need to factor this into your rate. You should go for "If I make this in 10 months per year, then it should be enough money". Also if you have several clients, you might feel bad about not being there for them. But, on the other hand, most of the people work with their employees, and it's normal that they just go on vacation. And that's another thing. If nobody stops you, you can just work 24x7. You have to take some breaks. (33:02)

Avoiding overworking

Alexey: So, if you have an hourly rate, you have this incentive of working all the time. You can get three clients and then work eight hours for each. And then, after a week, you burn out. How to avoid overworking. Do you have any tips? (33:58)

Mikio: I want to say it depends on the kind of work you do. If you're mentoring and working with people, you can't just work for eight hours per day. Maybe if you have 10 clients, you can slice them together every day. You have to keep an eye on your capacity. You have to find a balance — what makes sense for you in terms of money and go from that. (34:32)

Alexey: As a freelancer, now all of a sudden, you have a lot more things to worry about. You need to worry about setting my hourly rate. You don't just take the hourly rate that you make as a full-time employee and use it. You need to account for vacation, for sickness, for the risk that there will be no client. Is there any rule of thumb, like you can take your current hourly rate and double it or triple it? (35:10)

Mikio: I'm not sure. I don't know. (35:44)

Alexey: This seems quite difficult. So, you have this problem. On top of that, you also need to think about your mental health and avoid overworking — think how do you let yourself take a vacation? It's a lot to manage. (35:47)

Specializing: consulting as a product

Mikio: It is. You also need to think this: consulting is like a product. In the beginning, you're like, "It sounds interesting. It sounds like something I could do." You take on a lot of projects and every project is different. But with time, you want to standardise what you do and also specialise more. So maybe there's this one thing that you do. Like, certain kinds of models. (36:11)

Alexey: Like a niche, right? (36:56)

Mikio: With that, everything becomes more predictable. You get better at what you do. You always do certain kinds of projects. Every time you do it, you learn a bit more. And you get more predictability. You also get better at what you do. And your value for that grows. I think this is what many people do overtime. But that's something you have to think about it. And then also say "no" to some projects. "This is a lot of money, a very interesting client, but this is not what I want to do." Once a month, I'm like, "What am I doing here? Is this good or not?" And then you start thinking about it. But this is also a part of what makes this so interesting. (37:03)

Alexey: So when you agree to take a client, you don't know if you like it or not. There's only one way to find it out if this is what you like — to try it. By trial and error, you focus on a specific thing. (38:04)

Mikio: And then there are different kinds of situations. Maybe this is something you're good at, but it's not fun. Or it's a lot of fun, but you're really struggling with it. So you need to do a lot of exploration. (38:19)

Alexey: Maybe you need a bit of both? You need something fun, but you need to learn at the same time. (38:37)

Working full-time as a principal vs being a consultant

Alexey: You have multiple clients at the same time. What does your day look like? You work at the same time for them, during one day, you have meetings with multiple clients? Or one day — meetings for one client, and another day — for another client? (38:46)

Mikio: It's up to you how you want to organise it. But I manage a different Google calendar for each client and then try to update all the availabilities and keep them in sync. This reminds me of Zalando when I had a principal role. There you have different projects, and you meet with them during the day. That actually sounds pretty normal to me. (39:02)

Alexey: So how is your work now different from your work at Zalando? (39:32)

Mikio: It's similar. But I have more freedom. At Zalando, if there's nothing happening right now, you can't do anything. But now I can put together my own mix. It's very interesting to see all these companies, different company cultures, different approaches, different tech stacks, and so on. (39:35)

Alexey: Do you now work less than you worked full-time or more? (40:09)

Mikio: In the past few weeks, it seems like I'm working much more. In general, it was similar. (40:13)

Alexey: 40 hours per week, approximately? (40:23)

Mikio: Yeah. I do normal 10-to-6 days. But not all of that time is billed. The actual client hours make maybe up to half of it. And then there's other stuff like, like this [podcast], or working on talks, or just learning something. (40:24)

Alexey: In terms of salary, if it's not a secret, do you make more than working full-time or approximately the same amount? (40:33)

Mikio: So far it's approximately the same, but let's see where I end up at the end of the year. Some months it was definitely more. But I can't complain. (41:03)

Is the overhead worth it?

Alexey: Because if you're doing pretty much the same work, and you're getting the same money, is it worth all this overhead? (41:19)

Mikio: Yes. Now I have freedom and can control things. Also, we haven't talked about this, but when I started, I thought "Maybe I do a startup?". Or I work on some of my own ideas. At any point, you can say, "I don't take new clients", (41:28)

Alexey: Or "I'm taking only one client". (41:51)

Mikio: Yeah. I think it's much more interesting than what you can do within one company. (41:55)

Alexey: So if you work at a company, you have to devote all 40 hours to the company. But when you work as a freelancer, at some point, you can say, "I'm working no longer than 20 hours for my freelance projects. And I'm working the remaining 20 hours for my cool stuff." Maybe a new streaming analytics startup? (42:04)

Mikio: Yeah, there's still so much potential. But also there's something I had to really get used to. There's no boss now. You have to do good work for your clients, but then you can just take the afternoon off — if you manage to keep it free of meetings — and do your own stuff. It's all up to you. (42:28)

Alexey: You have more freedom. If you want to work from a different country, you just go and work. (43:00)

Mikio: Also, sometimes at a company — I will not say which one — there's not a lot to do, and you're left there for eight hours per day. And you feel like "I owe them this time". But now if that happens, you can just do a couple of meetings, and then spend the rest of the day doing whatever you want. (43:12)

Mikio: I feel a bit like my dentist. My dentist is just next door. And sometimes I see him in the street. I feel like "I'm the dentist now. When I work, I make money. If I don't, I don't. When I close my practice, I can go play golf." I don't play golf. But there's really a bit of being an entrepreneur in this. (43:30)

Alexey: So you get the freedom, but you need to worry about more things — also about promoting yourself. You have to do a lot of work that you don't bill, but you need to do it in order to get new clients. (44:03)

Getting a new client when you already have project

Alexey: Let's say you already have a few clients. And then there is a new client who comes to you and says, "I have this new project for you". You really like this project, this is what you really want to do. But you don't have enough capacity right now. What do you do in this case? (44:20)

Mikio: I don't think that's a good move, but you could say "I really want to do this. Let me ramp down on my other projects, and then I do this one." But in the end, it's up to you. They probably won't do another project with you again. But ultimately, maybe, doing this other project is better. (44:50)

After freelancing: what’s next?

Mikio: There was another book called "Built to Sell". If you build a startup, you have the plan of selling it. It's a lot about standardising your product. If you start as a consultant and you get lots of new projects. One option would be to found a company and hire a few people. Then you're trying to build up a consulting agency. This book says, if you want to sell it at some point, it has to run without you. You have to standardise everything. But that also involves saying "no" to clients that you had for a long time. This is also a part of what you can do. (45:15)

Alexey: What are your plans? Do you want to have a consultancy, or you want to have a start-up? Or you don't know yet? (46:09)

Mikio: I don't know yet. I'm not sure. (46:18)

Alexey: But in general, when you start freelancing, you don't want to do this until your retirement. What could be the next moves after freelancing? You can start a consultancy company, you can hire new people, they will help you work. You can also start a startup, if you see that some of the projects from clients repeat — they're about the same thing. You can put this in a product and then start a product company. (46:23)

Mikio: Yeah, there are so many options. Some just freelancing, consulting, and the freedom. You get better at negotiating than I am, and you make enough money so that your retirement is taken care of. You can do this for a while, and then you find a company — a great client — and then you think, "I'm done with experimenting, I can become an employee again." At Zalando, I interviewed a lot of people. Every now and then there was a consultant. When I ask them why, they always say that they want more ownership. I haven't experienced that yet, but if you do this for a few years, it can happen. Because you always do one part of the project, and then you move on. (47:03)

Mikio: Then, as you said, if you are more hands-on and you realise you're doing the same thing, you build a product that can do that. You can also build up a consulting agency. But then you're somebody's boss. It's very different from just being a freelancer. (47:59)

Alexey: Now all of a sudden you have less freedom. You have a company to run, you have to care about people. (48:20)

Output of Mikio’s work

Alexey: I remember seeing a question — What is the output of your work? Is it a report, a presentation? What do clients get at the end? (48:27)

Mikio: That depends a lot on what you do. In my case, I have more longer-running projects, like 3-6 months, where I work with the team. The output actually is what the team did in the end. As part of that, I'm doing concepts and writing that down. (48:50)

Alexey: Like a principal, right? This is what the principal would do at Zalando — they help a team do better. Right? (49:10)

Mikio: Yes. Also, if you go into productionizing more, you do workshops with a company. Then you do an analysis and say, "This is the stuff you need to work on". This could be an outcome. If you're a hands-on freelancer, then your code is the outcome. For data science, I know one guy who does a lot of PoCs. He goes to a company that has no data science capabilities. He discusses with them what they want to work on. They produce some data. And then he does the first prototype to see whether it's worth going deeper. If you like it, there's a lot of stuff you can do and find something that makes sense for you. (49:21)

Learning new things

Alexey: We talked a bit about that. There are things you're good at. And there are things you are not good at yet, but you want to learn them. If there is something interesting you want to be better at, how do you do that as a freelancer? Let's say, project management. You cannot tell your client "I actually suck at this, but I want to learn this thing for your money. Hire me." (50:17)

Mikio: Usually, there's a thing that you're good at, but you want to get better at it. You're looking for stretch assignments — stuff, where you think like "I can do this, but it's a bit difficult." You can also be open about it. You can say "I never did this before. I cannot guarantee whether it works. But let's try." Also, during the project, there are often opportunities — there is something that needs to be taken care of. Then you can try to do some of these things. You can also team up with other people, see how they do it and learn from them. And all the other classical stuff. (51:01)

Mikio: If you want to do this, you're looking for projects, which are a little bit outside of your comfort zone. But they shouldn't be so far that you won't be able to do it. (51:50)

Alexey: So you need to feel comfortable about being able to deliver the project. (52:07)

Mikio: Yes. (52:11)

Alexey: Not moon-shot, but something real. yeah. Like at work, maybe? Or at work, you have more leeway, because you can say, "I'll just learn whatever new technology". But you can't tell this to your client, and learn for the money of the client. (52:14)

Mikio: That's true. This is then your unpaid hours. But you have time to do that if you want to. (52:36)

Lessons learned after finding clients

Alexey: We have quite a few questions. This one is very interesting. What is your most relevant learning when finding clients? What did you learn from this process of finding clients? (52:45)

Mikio: Writing a summary is really helpful. I felt that this is an extra step. But actually, it's good to be really clear about what the project is before you start it. (52:57)

Alexey: So, make sure you're on the same page — in writing. (53:22)

Registering as a freelancer in Germany

Alexey: Then there is a question from somebody from Germany. And the question is "How do you deal with all this 'normal' German paperwork?" It's a lot of bureaucracy. You need to fill a lot of papers, you need to do this monthly tax declaration. How do you deal with all that? (53:30)

Mikio: I'm not a tax accountant, so don't sue me. In Germany, there's this "Freiberufler" concept. Your work is based on your expertise, like for architects, doctors, and so on. They have a special setup that doesn't exist in many countries. One good thing is that you just pay income tax. If you have a company, you have to pay tax on the revenue, but here it's just that [income]. Actually, it's quite simple, you write a letter to your tax office and you tell them you're a freelancer. They asked me what is the estimate for my income for next year, and based on that, they say how much you pay in advance. Then you get a tax number for this activity, which is different from your employee work. With that you can apply for VAT-something-id. (53:53)

Alexey: That's getting too technical. Do you do this yourself? Or do you have a tax advisor? (54:21)

Mikio: No, I don't. I should get one. But I thought I could try that myself. Also, based on the estimate, they tell you whether you have to do this monthly or in longer increments, and then you have to do at the beginning of the month. There's ELSTER.de where you can just do this stuff. It takes a while to figure out, but it's to just one or two numbers that you fill in. (55:25)

Alexey: If it takes a while for Germans to figure out how to use it, then foreigners are doomed. (55:55)

Mikio: It's not that difficult. You only need to know five things. Let's see what happens next, when I have to do my tax declaration. (55:59)

Alexey: It was a two-part question. The second part was if you decided to operate as a freelancer or as a company. I think you answered that — "as a freelancer". The alternative is to start a one-person company. You decided that for your case it's better to start as a freelancer? (56:06)

Mikio: Yes. If you start a company, that's a lot more hassle. You have to do it properly, you have to go to a Notar and do this funny thing where he reads you the whole contract, and then you sign it. I don't know whether you needed to do it before. Have you done that already? (56:13)

Alexey: Yes, for a different thing, for a mortgage. Yeah, it's 1.5 hours and the notary is just reading out loud, very fast. (56:47)

Personal liability of a freelancer

Mikio: Yes, that's how it's done. It takes forever. Then you have to do this yearly tax. You can't do this alone. But as a freelancer, you have personal liability. But there are professional insurances. They are quite expensive, a few hundred euros. They cover that, I hope. (57:02)

Alexey: So, technically a company can sue you if you do something that affects their business, right? (57:24)

Mikio: Yes. (57:30)

Alexey: That's why you need this insurance to make sure that if that happens, you don't have to sell everything you have just to pay for that. (57:32)

Mikio: It's the same for doctors and architects. (57:46)

Alexey: Would you recommend having this insurance? (57:50)

Mikio: Yes. Some clients also request that I have it. (57:53)

Alexey: Okay. So they will not work with you unless you have that. They want to make sure that if something happens to the production database, they are safe. (57:56)

Mikio: Yes, if I delete the production database. (58:07)

Alexey: How would you do this if you don't do hands-on work? (58:10)

Mikio: I don't, yes. There's also stuff like if you're onside and somebody trips over your laptop, and they get hurt — all this stuff. It covers a lot. There's GDPR and data. What if you lose the data? If you start reading insurance papers, you just wonder "There's a lot of stuff that can go wrong". (58:13)

Alexey: Probably insurance companies did witness all these things going wrong. (58:44)

Mikio: But actually becoming a freelancer and Germany is less complicated than you think. (58:50)

Effect of globalization and remote work on consulting

Alexey: Do you think the freelance market is affected by globalisation? Now everyone can work remotely. Why would a German company hire you instead of going to China or India or some other countries? The labor is cheaper there. (58:59)

Mikio: I guess yes. It also depends on what you want to do. If you're a front-end freelancer, it's commoditized. There are websites where you can go if you want to do a data science project. There, you're competing globally. But if you're more specialised or if you have a name, then that's how you get it — because everybody knows that it's super hard to find good data scientists. (59:22)

Mikio: It was also the reason why I started now. I always thought about it, but then I thought I could never travel that much. If you have a number of clients, you have to travel a lot. I didn't want to do that because of my family — small kids and so on. (1:00:14)

Alexey: If you have clients from different parts of Germany, or Europe, or the world, then you would need to go on-site and talk to them. Now, since everyone is doing things remotely, you don't have to do that. Right? (1:00:34)

Mikio: Yes. I talked to big consulting companies, they say, "things really have changed". People understood that remote work also works. (1:00:46)

Advice for people who want to start freelancing

Alexey: Before we wrap up, do you have any advice for people who want to start freelancing? (1:01:02)

Mikio: Yes. Just try. Mostly data scientists are watching this. right. People are really looking super hard to find good data scientists. The market is really in our favour. If you have a good network and if you know people who would hire you if you quit, then there's a good chance. If you have enough money to try it out for a few months, just do it. (1:01:10)

Woking full-time and freelancing at the same time

Alexey: What do you think about freelancing and working at a company full-time at the same time? Is it realistic? If somebody wants to find out if freelancer is for them. If they think "We don't want to quit our job and have no certainty whether it will work out or not?" (1:02:11)

Mikio: For me, it wasn't realistic. When I had a job, I had no energy left. Maybe if you're young... Also if you're already working full-time, it's hard to also work with the client during office hours. (1:02:35)

Alexey: Yeah, they want you to be present during office hours. So you need to find a client from the States or from Australia. And then you don't have a life [laughs]. (1:02:56)

Wrapping up

Alexey: How can people find you? (1:03:12)

Mikio: On LinkedIn. Also on Twitter, @mikiobraun. Follow me. I've been stuck with my followers count for two or three years now. (1:03:14)

Alexey: Please follow Mikio. (1:03:28)

Mikio: I need more followers. (1:03:29)

Alexey: Maybe if you make a couple of announcements about jblas, people will follow you. (1:03:31)

Mikio: Yeah, people are still using jblas, interestingly. (1:03:35)

Alexey: It's maybe a topic for another discussion. (1:03:43)

Mikio: It is. (1:03:46)

Alexey: Thanks for joining us today, for sharing your story with us, for giving us your tips. Thanks, everyone, for watching and for asking questions. Have a great weekend, everyone. (1:03:47)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.


DataTalks.Club. Hosted on GitHub Pages. We use cookies.