Building a Career in AI: From Real Estate to AI Engineering | Gustaf Gyllensporre

The transcripts are edited for clarity, sometimes with AI. If you notice any incorrect information, let us know.

AI Engineering Field Guide

Alexey: Hi everyone, welcome to our event. This event is brought to you by the DataTalksClub which is a community of people who love data. We have weekly events and today is one of such events. Actually today I think we have two. There will be another one just a few hours from now. (0:00)

Alexey: If you want to find out about the events we have in our schedule there is a link in the description. Click on that link and you will see all the events we have in our pipeline. Do not forget to subscribe to our YouTube channel. Right now actually I think we have a bit more than 55,000 subscribers. With your help I hope to get a silver button for 100,000. (0:16)

Alexey: I hope to get this soon. Please subscribe. I think there are 10,000 left. Last but not least, do not forget to join our Slack where you can hang out with other data enthusiasts. During today's interview, you can ask any question you want. (0:33)

Alexey: There is a pinned link in the live chat. Click on that link, ask your questions, and I will be covering these questions during the interview. Last interview I forgot to cover some of the questions you asked. Today I am going to keep an eye on these questions. Please use this slider link to ask questions. (0:51)

Alexey: I am going to stop sharing that. I am going to open the questions that we prepared. If we are ready we can start. (1:18)

Alexey: This week we will talk about building a career in AI. We will talk about switching to AI engineering. We have a very special guest today named Gustaf. There is an interesting story behind how we met. I do not know if you know that there is a repo called AI engineering field guide. (1:30)

Alexey: This is a repo I put together five months ago. The goal for this repo is to become a hub for all things related to AI engineering. We have job analysis there. We also have a list of awesome resources. I was doing a lot of analysis collecting a lot of interesting things online with the help of AI. (1:56)

Alexey: I found a few interesting videos. A few months after that, Gustaf reached out to me saying he noticed I included his video. He suggested we get to know each other. This is how we got to know each other. Thanks a lot Gustaf for reaching out. (2:28)

Alexey: It is amazing to be able to get to know you. Today we are going to talk about your story. You shared some things with me about your career. I got really excited about it. I thought I have to have you on the podcast. (2:48)

Alexey: Gustaf is a former real estate agent. This is what made it very interesting for me because he turned into a senior AI engineer. His frustrations with the lack of technology in the real estate industry prompted him to build his own. Since then he led the development of conversational search for e-commerce to depict AI. He is currently building the agentic operational system for construction at brianta. (3:01)

Alexey: He has a lot of open source contributions including CPython and langraph. That is very interesting. Welcome to the interview, Gustaf. (3:28)

Gustaf: Thank you. I am excited to be here. (3:41)

Alexey: I am excited about this too. I already mentioned that your career is interesting. I also did a bit of things that were not related to IT at all. Not every real estate agent becomes an AI engineer, do they? Can you tell us about your career journey so far? (3:47)

Gustaf: I think I can quite confidently say I am probably the only real estate agent that became an AI engineer. Real estate agents in general are not very technical people. Let alone becoming a software developer or an AI engineer, I am probably one of the few. (4:04)

Alexey: Maybe it is a downgrade for them. That is why they do not do this. (4:19)

Gustaf: Real estate agents are good at talking on the phone. They are great with people. The technology part is a bit different. I already forgot the question. What was the question? (4:25)

Alexey: The question is can you tell us about your career journey so far? You can start with your education because this is interesting. How did you get into being a real estate agent? How did you start doing what you do now? (4:36)

Self Taught AI Engineer Pivot

Gustaf: I am from Sweden. I ended up moving to Miami for university. I was a big computer nerd. I was just playing video games all the time. I realized there must be more to life. (5:01)

Gustaf: I had to learn how to touch grass and be more social. Moving to Miami felt like a good way of doing so. I ended up studying real estate because I thought I was quite analytical. I thought it was cool. It is kind of like finance but with something more tangible. (5:12)

Gustaf: I ended up becoming a real estate agent. This felt like a good career for me because it allowed me to challenge my social skills. I learned how to talk to people which I did quite well. In my first year I sold about three million dollars of real estate. That was quite great. (5:25)

Alexey: Why change careers then? You were already making good money. I do not think AI engineers can make that much. The cap is probably different for engineers, isn't it? (5:49)

Gustaf: The cap is unlimited but I am about to get into why I left. COVID happened causing a lot of interesting things. Mainly I was on a student visa that was expiring. I basically had to leave the country and move back to Sweden. Before we get there, in parallel while I was a real estate agent I was frustrated with getting a lot of data. (5:55)

Gustaf: Generating reports took a lot of time. I realized a computer needs to do this. I could not be doing this. I needed to be on the phones talking to people. I needed to do showings. (6:21)

Gustaf: After work I was staying up super late googling how to automate stuff in Python. Within a couple of months I was building stuff that would generate reports for me. (6:31)

Alexey: So you did not have any technical background before that. You were just playing computer games as you said basically, correct? (6:48)

Gustaf: I think that helped because I had been on a computer a lot. I just found it super fun. This was before generative AI so I had to build everything from scratch. For me it ticked the same boxes in my brain as playing video games. It was super fun. (6:53)

Gustaf: I would stay up late at night building stuff. I would go back to do my real estate job during the day. I squeezed in a couple of hours of coding whenever I could. Eventually I landed with a set of tools that really helped me and other people in the company. I started developing tools for other people in the company just because it was super fun. (7:10)

Gustaf: I started falling in love with coding. I realized this was super cool. I thought I was in a good position because I understood the business side of things in real estate as well as the technical side. As I said before a lot of people in real estate are not very technical. It ended up being good for me. (7:28)

Gustaf: I left to move back to Sweden and was presented with two choices. The first was real estate in Sweden where it is a lot more regulated. They basically told me I had to go back to school for at least two years. I decided I was not going back to school. I had already done so much in real estate so there was no way I was going back to school. (7:46)

Gustaf: The other choice was to take the sets of tools I built for real estate agents and turn them into a startup. I did this and called the startup Kono Cube. I did that for a couple of years and ended up getting some customers. Long story short, it was a solution very niched to the Miami market. I could not really live in Miami anymore but I was still able to get customers by visiting. (8:09)

Gustaf: It just was not very scalable. It did teach me a lot about software development and running a business. Since then I have been working as a software developer in a bunch of startups. These are mainly prop tech companies in the real estate niche. I also led conversational search at depict AI which is a company founded by Anton Usika who is now the founder of Lovable. (8:35)

Gustaf: Currently I am back in the construction industry at Brianta as the founding AI engineer. That is the long story of my career. (9:09)

Alexey: It is still pretty interesting even though it is not short. These stories that are not short are usually interesting. You have a lot of things to talk about. This is cool. I never even considered contributing to CPython. (9:21)

CPython Open Source Contributions

Alexey: How did it happen that you had to contribute as a technical person? What kind of problem did you see that prompted you to contribute there? Were you just looking to do something and checking if there were any open issues for CPython? Did you want to see if you could take any of them? (9:39)

Gustaf: It is a funny story actually. I went to PyCon in the US last May. After the conference they had a sprint where you can sit in a room with a bunch of the Python core developers. That was super cool meeting them. Basically they were just there to help you out. (10:07)

Gustaf: It was a session where anyone could dedicate time to improving the Python library itself. You could sit there, hunt down issues, and attack them. If you needed any help you could wave down one of the Python contributors which was super helpful. During that sprint I decided to tackle some issues that I found in the Python library. Honestly, finding an issue that was not already worked on by someone else was really hard. (10:24)

Gustaf: Finding an issue that you can actually comprehend and contribute to was the hard part. (10:53)

Alexey: That was in C, was it not? (10:53)

Gustaf: Yes, there are a lot of issues that are actually in the Python programming language itself. (11:01)

Alexey: Okay, like the standard library. I see. (11:07)

Gustaf: It also improved things there. I did a lot of work on the email standard library package. It was really cool because I got to talk to the core Python reviewers and have my code reviewed by them. It was very helpful. They gave me some really good insights on software development best practices. (11:14)

Gustaf: It was a super fun time. Now I have bragging rights of being able to say that I have contributed to the Python programming language. (11:37)

Alexey: It is really cool. Do you miss working as a real estate agent? (11:43)

Gustaf: Yes, definitely. (11:48)

Alexey: Okay. (11:53)

Gustaf: There are parts of it I really miss. It is fun to meet a lot of people. I felt in software development it can get a bit isolating because you are just stuck inside a computer a lot of the time. This is changing especially with the advent of forward deployment engineering. This is a mix between the two which is perfect and that is what I am doing right now. (11:53)

Gustaf: I am very passionate about buildings and real estate in general. Going to different buildings and meeting different people is great. There is a lot of stuff I do not miss. It is a give or take situation. (12:16)

Alexey: What I heard about the real estate market in Miami is that it is pretty lucrative to sell properties there, is that true? (12:34)

Gustaf: It is. One of the things I do not miss at all is having a really hard time taking time off work. For example I was not able to come back to Sweden for two or three years because the markets were so hot. I had to do showings on the weekends. I could never have a weekend away because I had showings at 9 PM. (12:40)

Gustaf: My schedule was completely open. My first ever sale was on Christmas Eve on the 24th. That is how little time I could take off. Literally Christmas is a big holiday for me, but that was the first sale and I had to do it. It was super fun and probably one of my favorite Christmases ever. (13:06)

Alexey: I guess when you live in Miami and your home is Sweden it is not an easy trip. You really have to plan for a few weeks. You cannot just decide on Friday that you are going to go. (13:28)

Tech YouTube Channel Growth

Alexey: You have this YouTube channel. Tell us more about this. (13:49)

Gustaf: Two years ago I started this YouTube channel called Prop Tech Founder. Prop Tech stands for real estate technology. Originally I was going to talk about being a founder and real estate technology. (13:53)

Alexey: When did you start it? What motivated you? (14:04)

Gustaf: One of the things is I wanted to challenge myself socially. Putting my face out there is something that was very scary. I have been on YouTube since 2008 making gaming videos. I never showed my face or used my voice because I was anonymous. (14:11)

Alexey: So you had a different YouTube channel? (14:28)

Gustaf: I have had a bunch. (14:34)

Alexey: So this was not new to you. (14:34)

Gustaf: It was more just gaming videos where I was not talking. I was always scared to put my face out there. I am at an age where I try to face anything that scares me. I knew I had an interesting story to share about a real estate agent turned software developer. That ended up being picked up by you because those videos about my transition were what people found the most interesting. (14:34)

Gustaf: Now it has evolved into a YouTube channel where I help people become AI engineers. A lot of people like hearing my perspective because I come from a sales background. For me it is kind of like if a real estate agent can become a senior AI engineer, what is stopping you? I try to share my unique insights. A lot of that comes from being someone on the business side of things. (15:06)

Gustaf: That is something I double down on. When I am an AI engineer I always think about how the code I am writing will save or generate more money. I think about things that many developers might not focus on. They might get focused on chasing the newest technology. I just hit three thousand subscribers yesterday, so it is super fun. (15:28)

Gustaf: I want to get more consistent with it. I have received a lot of comments saying it has helped them in their AI engineering career. That is super fun. (15:54)

Alexey: What is the name of the channel? I want to share it with everyone who is watching right now. (16:00)

Gustaf: It is Prop Tech Founder. (16:08)

Alexey: I will share it right now with everyone who is watching. (16:14)

Gustaf: If you are interested in preparing for AI engineer interviews, I have a whole playlist that is doing quite well. (16:20)

Alexey: Can we talk more about preparing for AI engineer interviews? This is how we met. My code discovered your channel eventually. In this AI engineering field guide I have, there is a very large section devoted to interviews. One of your videos was picked up by the coding agent and got listed. (16:27)

Alexey: Can you tell us more about what these videos are and talk about the actual interviews? (17:07)

Gustaf: I have a range of different videos on AI engineer interview preparation. They cover anything from getting your resume picked for the interview to common AI engineer interview questions. This is stuff that I have synthesized from interviewing for AI engineer jobs. I have learned a lot from there. It is also from working as an AI engineer. (17:12)

Gustaf: I have also been on the hiring side of things. I have been in charge of hiring AI engineers, so I have both perspectives. I try to share what I look for when I am recruiting AI engineers. (17:39)

AI Engineer Resume Optimization

Alexey: Let us take one of your videos and talk more about this. You have this unique perspective where you were looking for jobs, did a lot of analysis, and are hiring yourself. If I want to get hired right now and work as an AI engineer, what are my steps? How do I proceed especially if I am coming from a nontechnical background and am a self taught software engineer? (18:26)

Gustaf: Let us take it step by step. The first step is making a CV or resume that actually gets looked at by someone who finds it interesting. The key word here is interesting because so many resumes say pretty much the same thing. They have a tagline where they claim to be passionate about developing technology, but it says nothing. Probably because now it is so easy to use generative AI solutions. (18:43)

Alexey: It is very generic. (19:04)

Gustaf: That is part of it, but the other part is just that a lot of people do not do enough interesting things. When people learn software development, they all start making the same calculator or to do list apps. Unfortunately that is not enough anymore in this job market. You kind of need to stand out a little bit. That is the first step in figuring out how to make your background pop out. (19:09)

Gustaf: That is where you can get into things like contributing to Python or other open source projects. When I was interviewing for jobs, my tagline included that I was a contributor to CPython. That is something that not a lot of candidates have. It is an interesting conversation starter that makes someone want to talk to you. (19:36)

Alexey: That was exactly the conversation starter we had here because this brought my attention. I do not know anyone who contributed to CPython except you. (19:59)

Gustaf: I also had langraph in there because langraph and langchain are something that a lot of AI engineering companies are looking for. To be honest, it is less impressive for me because I do not find this tool very interesting. (20:06)

Alexey: I get your point. In this engineering field guide we analyzed a lot of job descriptions and langchain or langraph is the number one framework that people use. This is a very valid point. Most companies use it, so if you contribute to this framework, you stand out. CPython is cooler in my personal opinion. (20:19)

Gustaf: I think everyone would agree. We are taking the perspective of someone transitioning from a nontechnical background. This is a bit hard, but basically you have to build up the skills you need in your regional job market. You need to scrape AI engineer jobs and generate a list of technologies and concepts you need to learn. Of course you can get into projects and learn by doing them, which I think you should. (20:42)

Gustaf: The problem with projects is they are just not very interesting. If everyone is making a chatbot application with retrieval augmented generation, it does not tell me a lot about the candidate. What really unlocks opportunities is if you develop an actual solution to a personal or business problem you face. Maybe try to make a mini startup. That is what helped me both learn and stand out with my own startup Kono Cube. (21:15)

Gustaf: It was a personal project that ended up becoming so much more. It solved an actual problem that real estate agents faced. It actually had a couple of customers which is not necessary for getting a job, but it certainly helps. It makes you appear more interesting and brings you closer to understanding code in production. You learn how it translates to real value which is something the market desperately needs right now. (21:53)

Gustaf: We do not just need a bunch of people who can code. We need people who can turn code into actual business value. That is what we are looking for. (22:32)

AI Engineering Portfolio Projects

Alexey: I agree completely. I wanted to ask you more about this chatbot and retrieval augmented generation application. I agree that a lot of people have this, but a chatbot is not only about sending requests to the OpenAI API. There is already some curation work because you have to find a dataset and make your chatbot work with it. If I am interested in real estate regulations in Germany and have some domain knowledge, I could use that. (22:44)

Alexey: Maybe I worked in Germany as a real estate agent. If I use this knowledge to create a chatbot that other real estate agents can use because they are not technical, that is probably going to be very valuable, is it not? (23:17)

Gustaf: There you go. You are attaching a story to the chatbot. It is more than just building some technology because it is a complete solution. That is the interesting part. (23:33)

Alexey: The reason I am trying to find out more about this chatbot thing is because right now we have a course. The course is called LLM Zoom Camp. In this course we show how to integrate AI into products. The main focus of this course is retrieval augmented generation. Ninety five percent of people will graduate from this course with a project which is a retrieval augmented generation chatbot. (23:47)

Alexey: My question to you is how to make it interesting. All these people are watching our conversation right now and wondering what kind of recommendation you can give them. How can they make this chatbot more interesting or pivot from this idea to do something else? They still want to create a project they can use as a portfolio piece for AI engineering. (24:14)

Gustaf: I would start on the technical side without going towards the business side. Something I find interesting as an interviewer is if you have set up evaluations or thought about evaluations for the application. I find it super interesting because that is something that is very hard to do. There are so many ways to do it. If you even considered setting up an evaluation suite, I think that is quite interesting. (24:52)

Gustaf: The second part is taking this chatbot and attaching it to a real problem. A great way to do this is finding a dataset on Kaggle in real estate and seeing how it can be applied in the real world. You can go on Reddit and find problems that people are facing to tailor your chatbot towards solving that problem. This is nice when you are interviewing because it becomes part of the story you can tell. It is not just a chatbot you built because it is cool and you wanted to learn, but you actually saw a problem and decided to tackle it. (25:28)

Gustaf: It does not mean you need to have customers, it is just interesting that you found a problem and provided a technical solution. Those are the problem solvers that make desirable AI engineers. I would also think beyond just a classic chatbot application with vector search. If you can combine skills or file systems, I think that is very interesting. That is the direction where generative AI is going. (26:26)

Gustaf: We can have deep agents that take a lot longer to think and navigate through file trees. That is very interesting. That is what comes to the top of my mind. (26:58)

Alexey: We do not teach how to go deeper, but I would encourage everyone who is watching this course to go beyond the material. Find out what else you can do on top of that to make your project a little bit different and unique. Then you can actually stand out from the rest of the people who graduate from the same cohort. You probably do not even need to name your project a chatbot. Even though the interface at the end could be a chatbot, you can just say it is a real estate agent for ML engineers in Germany. (27:15)

Gustaf: Frame it more as a solution rather than just the technology. (28:04)

Alexey: Yes, and then the chatbot is just part of the interface indicating how the person interacts with it. It does not have to be called a chatbot. (28:04)

Gustaf: It is all about how you frame things. I have looked at CVs where it just says they built a conversational chatbot. That is not interesting. If you build an actual system or a mini OS for real estate agents, it is about what you build and how you frame it. (28:11)

Alexey: I see. Saying you built a conversational chatbot is like saying you built an app with Python and Django, or a to do list in React. (28:29)

Gustaf: Yes, but you kind of do not mention that. (28:40)

Alexey: You want to focus on the specific problem you solved with these technologies rather than starting with the technologies themselves. (28:46)

Gustaf: As a disclaimer, my career is very rooted in startups and I have only been an interviewer in startups. It might look different in bigger companies, but these are the type of people we are looking for. We want problem solvers applying technology to real business value. I am very biased towards that perspective, but I also think startups are the best way to break into this industry, especially if you are a nontechnical person. (28:53)

Startup vs Big Tech Interviews

Alexey: Speaking of the interview process, startups focus less on things that big tech is focusing on, like LeetCode and system design. You probably know if you look up interviews with Meta, Amazon, or Google, you will see a very structured process. It is always a screen coding session with medium LeetCode problems. For me with Meta I had two screening interviews and then an onsite which was online. I had two LeetCode interviews, system design, machine learning system design, and behavioral questions. (29:33)

Alexey: You can find all these things online, but my experience interviewing with startups is they do not really ask these things. The focus is a bit different. Does your experience being on the interviewee side confirm this? (30:20)

Gustaf: Absolutely. Bigger tech companies and enterprises all have a structured interview process which breeds a culture of conformity. If you are an oddball like me, you do not really fit in there because I never wanted to grind LeetCode. Startups are more rewarding for people that stick out a little bit. That is why I am making all these points about making your profile and CV interesting. (30:42)

Gustaf: We focus less on credentials like where you went to university or if you studied computer science. It is more about what type of person you are and your approach to problem solving. It is about your passion, your energy, your vibe, and what you do in your free time. We want to know what you were doing when you were ten years younger. Were you creating Minecraft servers and hosting them? (31:21)

Gustaf: That tells me a lot more about the person and who they can become rather than who they are today. (31:50)

Alexey: What did you do ten years ago? (31:56)

Gustaf: I was modding Minecraft. (32:02)

Alexey: So you had to learn some Java for that, didn't you? (32:02)

Gustaf: I did not learn anything, I just installed mods online. I did not know what I was doing half the time, but it was super fun. I did not learn any Java though. (32:09)

Alexey: I heard from many people that this is how they started programming. They were playing Minecraft and needed to do something, so they had to install Java and change some code. Because they got curious, they eventually started doing it more and decided to pursue software engineering as their career. Since Minecraft is quite old, we now have a generation of developers who discovered it as kids and now work in companies. (32:23)

Open Source AI Project Ideas

Alexey: I see a few questions from our audience. Someone is asking you to recommend a few projects because a generic chatbot application sounds boring. You mentioned that you want to make your CV stand out. What kind of projects make your CV stand out? Do you have a structured approach that you recommend to use when finding projects to make? (33:47)

Gustaf: In terms of a concrete project that stands out, one that is interesting to me is an AI hedge fund. It is a continuously developing open source project which has you impersonate a main investor like Warren Buffett. It uses different generative AI practices to come up with an investment strategy for a stock portfolio. I think it is super interesting because it uses a lot of different concepts and ties it to real business value. It is all open source and very easy to follow, making it a great project to start with. (33:55)

Gustaf: The best way is always to find a problem that you or other people are facing and try to tackle that. (34:34)

Alexey: You mentioned spending some time on Reddit and seeing what kind of problems people have. (34:45)

Gustaf: Exactly. For example, I have an open source project that is not very generative AI related at all, but I started it a couple of years ago due to a personal frustration. I was subscribing to Audible and you can click on a bookmark if there is a section you find interesting. Then there is no user interface for looking at the bookmarks. I created a program that extracts these bookmarks, transcribes them with AI, and exports them into Notion so I have a database of all my bookmarks. (34:52)

Alexey: Wait, so you listen to an Audible book, add a bookmark, and your app extracts what was discussed, creates a summary, and puts this to Notion? (35:33)

Gustaf: Yes, it transcribes that part of the audiobook into text. (35:50)

Alexey: That is so amazing. (35:50)

Gustaf: I made it a couple of years ago and posted it on Reddit in case anyone else had the same problem. It is quite active with many stars and a lot of people using it. Those are the type of projects that really make you stand out. It teaches you a lot of things because there was a problem you actually had. (35:56)

Alexey: You leave a bookmark and then you kind of have to go back and relisten. In my case I never do this. I have these bookmarks and if I remember something from the book that is good, but if I do not, I will not go back. Having a system that would let me go back and check it is cool. You had a personal problem, you solved it, and other people also found it useful. (36:14)

Gustaf: Exactly. I am thinking about all the different layers that this can teach an interviewer about me. First of all, this guy reads books, which is cool. He was frustrated by a problem and instead of just dealing with it, he implemented a solution. He was willing to share that solution to help others. (36:44)

Gustaf: I have other contributors on the project, so it shows I can manage code submissions from other people. There are issues that I have managed. This small project that does not make me any money tells a lot about my personality. In terms of something a bit more AI centric, it is hard to say because the best practices of generative AI are changing all the time. (37:08)

Gustaf: It is very hard for me to land on a concrete thing. Right now the trend is longer running deep agents. Latency is not really a problem anymore, whereas before the trend was chatbots answering questions quickly. Now it is more about building a long running agent that can solve very complex problems given a dataset that you index and chunk in a certain way. (37:48)

Building Deep Research AI Agents

Alexey: For example, one of the problems I have is dealing with five years of podcast transcripts. I interviewed a lot of people and there is a ton of useful information there. On average every episode has many interesting things. Let us say I want to write an article about getting started with AI engineering and I have already interviewed fifteen people about that. A deep research agent that would go through all these podcast episodes, understand which are relevant, extract interesting points, and summarize them into an article would be very useful. (38:55)

Alexey: I can do something like that if I just open Codex or Claude because all the transcripts are already in GitHub. I can ask it to write me an article, but it is not super interactive. Something like a deep research agent could be super useful. (39:26)

Gustaf: You just reminded me of a great project idea, the LLM wiki by Andre Karpathy. It is a system that takes a raw dataset of variant topics and condenses it into a wiki. You can make a personal wiki for yourself where it extracts patterns, entities, and concepts from your personal projects and life. You can make this second brain that can answer a lot of questions for you. I have a personal one and one of these at work. (39:44)

Gustaf: It is very interesting what you can extract from here. If people are finding it difficult to find a personal problem to attack, I think this is a very good starter. As long as you have notes or personal material, you can find some really interesting stuff about yourself. You can take it a step further and use an open source model like Ollama to synthesize it. (40:33)

Gustaf: If you have personal data you might not want to send it to Anthropic, so you can self host. Then you will get into prompt engineering and pattern extraction. I think that is very interesting and useful to know. (41:07)

Alexey: That is right. One of the things I keep mentioning in the course is that your retrieval augmented generation agent is only as good as your data is. The structuring part, when you take raw data and structure it, will make your application much more useful later. Otherwise it will just extract a bunch of chunks and the quality will probably not be so great. If you create a wiki, you have connections between articles and a linked graph structure. (41:20)

Gustaf: Yes, and you can build a whole application on top of that database and graph. There are a lot of different approaches here. You can have a file system retrieval which navigates through different folders and tries to find the relevant concept. There are a lot of cool things you can do that translate very well into being an AI engineer. (42:11)

Landing Your First AI Job

Alexey: Someone is asking how you mastered AI and machine learning. Could you share a list of projects you worked on? I think you already mentioned a few projects, like the real estate project that eventually became a startup. Along the way you probably did a few other things, so maybe you can just let us know about that. (42:52)

Gustaf: I am not sure I can claim to have mastered AI and machine learning. My introduction to AI was the machine learning specialization by DeepLearning.AI. I think it was very interesting, although honestly I do not use a lot of the things I learned in my job at all. It was useful because it really got me interested in AI and taught me what powers generative AI. I learned the low level stuff like back propagation and forward propagation. (43:19)

Gustaf: It gave me a nice certificate I could put on my CV and LinkedIn from a reputable source, since Andrew Ng is a very big figure in AI. After that I started doing my first project, which was a chatbot that I built on top of my own startup back in 2023. LangChain was just popping off, so I made a video about it and put it on YouTube. It got a couple thousand views, which was really cool and prompted me to do a course on it. (43:53)

Gustaf: I put the course that I made for this project on YouTube, and it got seven thousand views. It is an hour and a half long video about building a real estate agent using LangChain and vector databases. It was very well received and gave me something else to put on my resume. That ended up getting me my first AI engineer job offer. (44:41)

Alexey: How did it get you your first job? Did someone notice your course and reach out asking if you wanted to join the company? (45:18)

Gustaf: No, I actually just put it on my CV at the bottom under skills or things about me. I wrote that I made a YouTube course about LangChain and it had this amount of views. That is just very interesting, especially since the company I was interviewing with was running on LangChain. For them it was a no brainer to want to talk to me. (45:30)

Alexey: And they hired you after it. (45:55)

Gustaf: Yes. A very good concept here is that beyond just learning technical skills, content creation can be a very useful tool if you are trying to land a job. It does not have to be making YouTube videos, posting on X is also an option. There are a lot of people who get hired purely off of their X profiles. YouTube videos have the added benefit of showing immediately that you can communicate clearly and teach things. (46:01)

Tech Networking Strategies

Gustaf: An email newsletter will also help you in your learning journey, so I would not discount it. Another key insight I have learned about building a career in the AI engineering space is ambassador programs. I am an ambassador for LangChain in Sweden, so I get a budget from them and host events for people that build with AI. I do not even have to push LangChain. The amount of connections and job opportunities I have gotten through that has been insane. (46:48)

Gustaf: There are a lot of companies doing these ambassador programs, like Cursor and Lovable, and it is a win win. It is a win for the company because you spread their brand, and it is a win for you because it puts you in a position of authority. Even though you might not be an AI engineer yet, you are still the Cursor guy or the LangChain guy. People put you on this point of authority and find you interesting, which can lead to great job opportunities. (47:18)

Alexey: That is really cool. Two things stand out: content creation and checking tools you use to see if they have brand ambassador programs. If I wanted to start a YouTube channel and post videos, how would you recommend I do that? What do I need, a fancy microphone and camera? (47:51)

Gustaf: No, you can start with your phone. (48:26)

Alexey: I did not expect to hear a phone. I thought you would just say use your laptop and record a video of yourself. (48:32)

Gustaf: I was using a Linux laptop for the longest time and the webcam was terrible, so I had to use my phone. I would say keep it as simple as possible and definitely do not try to be a perfectionist. For a lot of my videos I think about a topic, start recording, and then talk for a long time. Then I go into my editing software and trim out the unnecessary parts. I tried to really overengineer my videos before and it took a lot of time. (48:38)

Gustaf: For example, that LangChain course is a ninety minute video that took so long to make, but it is one of my worst performing videos. My highest performing videos are where I just have a list of bullet points and talk about them. You will never know for sure what your channel is going to be in the beginning, so just get started. Put stuff out there, try different things, and eventually you will find something that sticks. (49:16)

Alexey: I am not a successful YouTube channel yet, I do not even have three thousand subscribers. Wait, in this channel it is almost ninety thousand, but I wish every video would get this amount of views. Seven thousand is a very good amount of views. Not everyone actually finishes, but it is a good proxy metric to understand how successful the video is. YouTube starts recommending it if it thinks it is a valuable video. (49:46)

Alexey: If they watch longer than average it continues to recommend it, which means it is a very good video. I wish many videos on this channel would be that good. The number of subscribers is a good thing, but it does not guarantee you a good amount of views. Let us say I am working on a course project and I want to record a video about how I made it. What do I need for that? (50:28)

Technical Project Demo Videos

Alexey: I probably can record a short demo or have a walkthrough of how I actually created it. How would you approach option A and option B? (51:16)

Gustaf: The first thing I would do is show what you actually built in the first couple of seconds. The first seconds are crucial for people to decide whether to stay on the video or click off. Show the project, make sure it is fancy, and captivate attention. Then I would go into the code, but not writing out the code. I would basically just show a picture or a video snippet of each important concept. (51:26)

Gustaf: You can highlight the piece of code and explain what it does. I would just tie the code to the actual project by explaining the code representation of each feature. You also find an added benefit in explaining things to people. Explaining stuff to people will really deepen your own understanding of the concept you are teaching. It will really help your own learning and journey to become an AI engineer, which is why content creation is so nice. (52:03)

Alexey: Most people do not do this. It is important in this day and age to stand out, especially if we talk about the AI engineering market. You want to stand out and this is one of the things you can do. (52:43)

Gustaf: We have not even covered having a properly optimized LinkedIn profile. A lot of people have dead LinkedIn profiles, which makes it seem like they are not an actual person. Unfortunately, that is something you need right now. You also need to know how to behave in interviews. There is this whole realm of things to think about. (53:09)

Gustaf: From someone who has interviewed a lot of candidates, a lot of people are just doing the same thing. The barrier to entry or impact is not really that high. It is just about approaching things a bit differently. (53:26)

Alexey: I see a very interesting question. The question is what mistakes do self taught AI engineers make that slow them down. You probably cannot speak on behalf of all AI engineers, but maybe you can reflect on your own journey. What kind of mistakes did you make that, if avoided, would have made you move faster? (53:49)

Gustaf: Mistakes, I am not sure I made any. I made plenty. I do mentoring for people that have AI engineer interviews coming up. I get exposure to a lot of people's frustrations and how they approach landing the job. I have helped a lot of my subscribers get AI engineer jobs through coaching, and I see a lot of patterns. (54:14)

Self Taught Developer Mistakes

Gustaf: People have seen this glamorized tech career where they want the remote job and all the perks. That does not come until later. Your first priority should be to get into the industry and get that job. If that job is in person and you have a long commute, that is what you have to do. In person jobs are severely underrated for career growth. (54:53)

Gustaf: Another mistake is trying to learn everything at once and being overwhelmed, which is easy in AI because new stuff drops all the time. The real approach is to look in your local area for the jobs you want to apply for. Ideally they are on site because that means less competition. Scrape the job listings and make a reasonable list of what you really need to learn. Commit to learning only those things and learn by doing projects. (55:16)

Gustaf: The last and really most important part is neglecting the social aspect of things. Back in the day, just knowing how to code was enough to get you a very high paying job. Now knowing how to code is nothing because coding agents do that anyway. What people are looking for, myself included, is people that can translate technology into business solutions and interact well with human beings. The technological part is solved on a low level by coding agents, so you need to understand customer needs and develop realistic solutions. (55:56)

Gustaf: When you are developing a feature, can you reason about how a customer or user would use it? That will guide you into a certain approach for developing the feature. Coming from a startup background, we place a lot of emphasis on whether a person is a good communicator and operates well in a team. It comes down to social skills. You need to be able to own a certain approach, think about its pros and cons, and be able to defend it. (56:37)

Gustaf: A certain degree of confidence is very well received. Having the confidence to flag stuff that you do not agree with is important. If you disagree with something, we want to hear it because nobody is an expert. (57:23)

Alexey: Agreeableness is something for AI. You are absolutely right. I wish it would be less agreeable and just tell me my idea is bad. (57:46)

Gustaf: We need more emphasis on the human side of things. At the end of the day you are interviewing a human. You do not need another robot, we have enough robots already. (57:59)

Alexey: Do you have to go now or do you have time for another question? (58:14)

Gustaf: I can stick around a couple of minutes more. (58:20)

Alexey: Let us keep it short. You mentioned the importance of social skills many times. What is one single thing people can do to improve that? We all know how to talk to an agent, you just type and the agent does something. (58:20)

Soft Skills for Software Engineers

Alexey: Talking to humans is much more difficult for an engineer. I have not worked in real estate, and recommending someone to work as a real estate agent is a bit far fetched. Do you have one single thing you would recommend engineers do to improve their social skills? (58:53)

Gustaf: I have an answer for this. It is something I did even after I was a real estate agent because I still had social anxiety. I do not think people are going to like this answer, but it is basically going out and talking to strangers. Go out on the street and if you see someone with cool shoes, ask where they got them. Try to start conversations with people. (59:06)

Gustaf: It was a scary thing here in Sweden because nobody does it and we keep to ourselves. However, you will find that it alleviates your social anxiety. It makes you more comfortable and fluid in talking about a wide range of different topics. When it comes to social skills, a lot of things are just about removing or limiting your social anxiety. It is not going to get removed by itself, you just have to break it down and decide you are going to talk to someone today. (59:28)

Gustaf: A slightly less hardcore version is going to a meetup or conference and talking to people there. (59:58)

Alexey: I live in Germany, and if I approach somebody on the street and ask about their shoes, they would look at me like I am crazy. Maybe it is my mental barrier. (1:00:04)

Gustaf: It is a mental barrier. Going to meetups and being an ambassador will force you to talk to people, which is a more natural way of doing it. A lot of people think that even if you live in a cold country like Sweden or Germany, people are going to think you are weird. They probably will think you are weird in the first couple of seconds, but eventually they will appreciate getting approached because it does not happen often. Attending events is a great way to start, and any way you can talk to new people will increase your social skills. (1:00:22)

Alexey: You mentioned mentoring and coaching. Where can we find more? Is it on your LinkedIn or somewhere else? (1:01:03)

Gustaf: It is on my YouTube. (1:01:09)

Alexey: I already shared the link to your YouTube channel. If anyone is interested in talking to you about their career challenges, they will find it on YouTube, is that correct? (1:01:15)

Gustaf: Yes, any description on any YouTube video will have my links to mentorship. (1:01:22)

Alexey: Thanks a lot, Gustaf. I really enjoyed this conversation because it was packed with so many useful things. I am really happy that we got to know each other and you managed to join this channel. If there is anything I can help you with, I will be glad to. Thanks everyone for joining us today and for listening in. (1:01:35)

Gustaf: Likewise, it was super fun. It is always great to connect with another person trying to educate in the AI engineering field. Hopefully my answers gave some value to people. (1:01:48)

Alexey: I am pretty sure they did. I am looking forward to staying connected with you and your viewers. Have a great week. I hope it will be less hot than the last week, at least in Europe. Stay cool and see you. (1:01:59)