How to Become an AI Engineer After a Career Break | Revathy Ramalingam

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

Why Move to AI? Using ChatGPT to Plan a Career Pivot

Alexey: Hi everyone. This week we will talk about restarting a tech career after a long break and the AI engineer role. Today's guest is Revathy. Do I pronounce it correctly? (0.0)

Revathy: Revathy. (12.0)

Alexey: Hi everyone. This week we will talk about restarting a tech career after a long break and the AI engineer role. Today the guest is Revathy, an AI engineer at a healthcare startup. She brings over nine years of experience in software engineering, originally in the telecom domain. (17.0)

Alexey: She is one of the students of our ML Zoomcamp. She also took the AI Dev Tools bootcamp and recently a course called AI Hero, which is not officially a part of the courses we have at DataTalks.Club but is still one of the courses I run. In this podcast, I am going to talk about her experience. She will share details with us on how she restarted her career, prepared for the AI engineering role, and managed to get back to the industry after a break. (31.0)

Alexey: Welcome. I am really excited to have you here because you are quite an active member of the community. Ever since our Machine Learning Zoomcamp started, I saw that you commented everywhere. It is always nice for me personally as someone who oversees the community manager role to see people like you joining and being active in helping others. (1:08)

Revathy: Thank you, Alexey. (1:48)

Alexey: I want to start by asking you to describe your career journey so far. Can you share with us your background and what you did before becoming an AI engineer? (1:54)

Revathy: I was a software architect at Ericsson, a leading telecom company. I worked there for about 9.3 years. I started my journey in Ericsson as a graduate trainee and then progressed to a senior software engineer. Finally, I was a software architect at the time of leaving the company. (2:07)

Revathy: Due to a maternity break, I could not continue working because of my kids. I had to take a break for about seven years. My elder son is thirteen years old and my younger son is eight years old. (2:26)

Revathy: I wanted to see them growing. With this fast paced software industry, you are not able to be with the kids because you are always under pressure. I wanted to have a break to take care of them and then progress later. (2:52)

Alexey: But then the career break became quite long. It was seven years. Currently, you are working as an AI engineer in a healthcare startup. What did you do in Ericsson? Was it Java development or something else? (3:16)

Revathy: I started as a C++ developer mostly on protocols, voice charging, and data charging. For four years, I was in C++ mainly on protocols. In the second half, I moved into Java projects, big data systems, and Apache Karaf. Camel architecture and Java REST services were some of the technologies that I worked with. (3:29)

Alexey: What are these things? Apache Karaf? (4:01)

Revathy: Apache Karaf is more or less like a Spring Boot service. You have all the bundles and you will be able to do a hot deployment. Just the framework is different. (4:01)

Alexey: I come from the Java background myself, so I was a Java developer at some point. I know what Spring Boot is. Can you tell us a few words about this? What exactly is Spring Boot and what is Karaf? Why would a Java developer care about these things? (4:12)

Revathy: I worked only with Karaf and Apache Camel architecture. This is a tool for creating web services. I have not worked with Spring Boot myself. (4:31)

Alexey: I only work with Spring Boot, but basically anytime you want to start a Spring application, Spring Boot is the fastest way. It bootstraps your application and then it is ready to start a web application in Spring and Java. (4:49)

Revathy: You can ship your products easily using Spring Boot. It was very much in demand in the current industry. (5:12)

Alexey: Why did you decide to return? (5:18)

Revathy: I was feeling bored at a point because my kids all went to school full time. I was on my job hunt because I was very determined to work. Right from my early career, I was a topper at school and a university rank holder in my graduation. I could not be idle at home and wanted to get back to my line of work. (5:18)

Revathy: Initially, I started without any upskilling. Within two months, I realized that because I was working on VMware for deployment, and now the industry has moved toward Docker and Kubernetes, the technology had changed so much. I soon realized that with my current skills, I would not be able to get a job. (5:45)

Revathy: One fine day, I checked in ChatGPT. I told it I am a person with nine years of experience and have a career break for seven years, and now I want to get back into the industry. I asked it to tell me the goal and the path by which I have to reach my destination. It gave me two options. (6:15)

Revathy: One was to master Spring Boot as a Java architect. The other way would be to go into machine learning. I became very curious because ChatGPT has immense knowledge and is able to give me answers in whatever language I ask. I was very curious about how it works. (6:33)

Revathy: I decided to choose the machine learning line. When I checked on that, it suggested DataTalks.Club and a cohort that was about to start. Initially, I had no clues about that; I just clicked on the link it gave and enrolled. The next day, I got an email from you with details about the batch. (7:04)

Revathy: I joined the Slack community. By August, even though the cohort started in September, I was already seeing the GitHub repository to see how things were working. Your videos started from the very basics. I am someone who is very much interested in mathematics, so vectors and those concepts were easy for me. (7:36)

Revathy: With nine years of experience, setting up things was not difficult. The Jupyter Notebook was very easy. Whatever vectors I input, I got immediate output. I became more and more engaged with the course as it progressed. (8:02)

Alexey: Interesting. So you are one of those people who decided to start even before the course officially started? (8:17)

Revathy: Yes, I started early. (8:23)

Alexey: It is good because all the materials are already there. My next question was what led you to transitioning into AI? From what I understood, it was pure chance and curiosity. (8:23)

Revathy: It was pure chance. But I had an interest and a passion to know how it works. There was curiosity that led me. Even with Java Spring Boot, I followed a tutorial and developed my own commerce transaction application, but it was not that interesting. It did not push me to the next level. (8:46)

Alexey: In the sense that it was very similar to what you were doing before the break? It is Java, while this is very new. (9:13)

Revathy: Yes, it was more monotonous. There was nothing new for me and it was not engaging. (9:22)

Alexey: I am really curious. You said that since you already have a lot of experience, setting up things was not difficult. You could set up Jupyter Notebook easily and follow the GitHub repository. Did you try to keep yourself up to date during these seven years? Did you try to do any pet projects? I imagine if you just spend time with kids, some skills might become rusty. (9:36)

Revathy: No, I was very happy having leisure time. I was enjoying my life and did not think about my career at all. Around the beginning of 2025, when I attended a couple of interviews, they were asking about Docker and Kubernetes. I enrolled in an online platform and fully completed the Docker course. (10:08)

Revathy: In that same platform, there was also machine learning. I watched about five minutes of the video, but they jumped into a huge line of code and I could not understand anything. With that, I came out of that tutorial. (10:35)

Alexey: Eventually, you continued interviewing and realized you needed to upskill yourself. That led to discovering DataTalks.Club. How was the course? (10:47)

Learning in Public: The Power of Community Support

Revathy: It was amazing. It is the best course that is happening for free. All the videos are from the basics, which is very good even for someone who does not know about machine learning. (11:00)

Revathy: The community support has been amazing throughout whenever I had any doubts. The way it is structured is also very nice with a video, a module, and then a homework assignment that checks your understanding. Asking us to post in public was another huge plus in my career. (11:22)

Revathy: I started to understand things better. A lot of prospective people also commented on my projects. When I did a project in telecom, a person from Nokia commented and asked me queries about my project. That was pretty interesting. (11:43)

Revathy: Then I felt that I was not doing a bad job and was actually good. It motivated me to progress. (12:03)

Alexey: Community support comes from people like you who are enthusiastic about what they learn and are always willing to share with others. Maybe there are a few people like you who help each other. Sometimes I go to Slack and see that all the questions people had were already answered by the community. (12:10)

Alexey: I feel like I do not need to do anything. That is really cool when you help each other. Maybe you understood eighty percent of things and somebody else understood a different eighty percent. You can just help each other and this is really amazing. (12:45)

Revathy: It took me back to my college days where we shared and learned things together. For someone who was away from a career, this community engagement was very helpful to get back into the line. When people share tutorials and even GitHub workflows, it helps. Someone from the group pinged me about that, and then I searched in ChatGPT. (12:56)

Revathy: I learned that we have all these technologies available. My learning curve was getting better after getting more actively engaged in the community. (13:29)

Alexey: I want to thank you personally for being a very active member of the community. Because of people like you, others say the community support is amazing. I really appreciate it. You mentioned you like the structure, the homework, and the nudge to share things in public. (13:42)

Alexey: You shared about your project in telecom and somebody from Nokia commented. Can you tell us more about that project? Was it a midterm project? (14:18)

Revathy: It was my capstone project. It was about predicting the network slice in networks. In everyday life, we watch TV or YouTube videos, which is a mobile broadband thing for which you do not need latency. It has to be uninterrupted. (14:25)

Revathy: If it is delayed, the person gets bored and leaves. For certain cases, like when you are doing a surgery where the latency should be very minimal or a satellite launch where zero latency is acceptable, it is different. There is a third category called IoT sensors where devices are deployed and give readings about temperature or anything else. (14:50)

Revathy: For those IoT devices, latency is accepted. Even if it sends data after two or three days, that is fine. (15:22)

Telecom Capstone: Predicting Network Slices with ML

Revathy: My project was aiming at predicting the classification type precisely. Accordingly, the network will be allocating the bandwidth to the users. Allocating the bandwidth is a very critical and important task for an optimal network to be given to the users. (15:37)

Revathy: I understand that this is based on the background I already have. It was easy for me to progress, but as I deployed it, I faced a lot of challenges. I got a dataset from forty two base stations on Kaggle. The data was generated using synthetic generators. (16:00)

Revathy: It was a real time dataset, and when I first got it, it was giving me full accuracy. When I checked in ChatGPT, it explained label decoding. In a network, there are certain parameters that appear very constantly. With those things, the machine learning is able to memorize the pattern and give full accuracy. (16:27)

Revathy: That is not what we need. I had to remove all those parameters and then rebuild it. I had very good learning in that and it was quite challenging. As I was progressing in my course, I was also checking with ChatGPT on how I would get my job. (16:46)

Revathy: I was always looking to get back into Ericsson. Related to that, I decided to do a project in telecom and that is how it went about. (17:06)

Alexey: You mentioned using ChatGPT quite a few times. Now I am curious exactly how you use it. You already mentioned looking for things to learn next. Do you just talk to it or are there some tricks to make it effective? (17:24)

Revathy: I was a person starting from zero and it was just like a conversation with ChatGPT. Nothing special. I guess in this case it all happened by luck. Even the DTC course happened by luck. (18:15)

Alexey: You also did another course about AI Dev Tools. I was going to ask how you managed to do multiple courses at the same time. You must have been excited about three courses at once. (18:38)

Revathy: At one point, I was feeling very burnt out and wondered why I took all three. But it gave me very good results and was fruitful at the end. I am very thankful for that. During the course launch for AI Dev Tools, I did ask you whether I could take it up because I was curious and interested in machine learning. (19:04)

Revathy: I was very skeptical because I had no knowledge of coding agents or the syllabus topics like MCPs. I had no clues about that. At the course launch, I asked you about it, and you replied that it would be very useful for me in machine learning. You said I could use those agents to easily build my products. (19:24)

Revathy: I believed in your words and enrolled in that too because I had a lot of free time. I wanted to get as much learning as possible and it was also free. (19:47)

Alexey: How did you manage the time? Towards the end of ML Zoomcamp, it is not the easiest thing because we have complex things like neural networks, Kubernetes, and serverless. These are not super easy things. At the same time, you were doing AI Dev Tools. How did you allocate time? (20:00)

Revathy: Every day, I stayed up until 3:00 AM after I put my kids to sleep. In one way, the community would be available by then because after 12:00, you would all be there. It was easy for me to ping and get my doubts clarified. In deep learning, I struggled a lot. (20:34)

Revathy: I was pretty sure my answer was wrong, but I did not know where it went wrong. People shared a lot of tutorials. I was very poor in visualization skills regarding how an image comes in and gets transformed pixel by pixel. Those logics were not very clear to me. (20:52)

Revathy: Then there was a very good YouTube material shared in the community showing how the image undergoes all the transformations. It was quite interesting to see, but to do the homework and put it into practice had a lot of challenges. With my experience, I was able to figure out where I went wrong. (21:09)

Alexey: That is good. (21:31)

Revathy: I knew my output was wrong and it was a good learning for me. There was a point where I was actually blocked. (21:31)

Alexey: To me, it sounds like you are the perfect target audience for the course. We want to attract developers who want to learn machine learning but do not know anything about it. It is good that ChatGPT suggested it to you. Tell me more about the AI Dev Tools course. (21:42)

"Vibe Coding" & Building Prototypes with AI Dev Tools

Revathy: I got more interested in that because it was vibe coding. Everything you ask for is in the prompt. You start with Lovable and you create a UI and then you proceed. Things were quite easy. (22:15)

Revathy: Towards the end when I did the project, I got a lot of confidence. With just an idea in mind, we give a prompt to create a front end, create a back end, and then create a database to integrate these things. I developed a project called Vigilance AI in my AI Dev Tools course. When I was able to realize it, I became very happy. (22:31)

Revathy: I am able to convert my idea into a reality. The course has empowered me so much that I am able to have a workable project in hand. Definitely, that course is very empowering and useful. (22:57)

Alexey: With all these tools, it is so cool that you can go from an idea to something concrete in one or two evenings. You have a good prototype that you can show to people, and it is actually working. (23:18)

Revathy: Yes, it is your idea, so you have the ownership and responsibility. It is a very proud moment getting it converted into a project. (23:32)

Alexey: Then you also did my personal course, AI Hero. This course is not a part of DataTalks.Club, but is a course I do through emails. I know you are still waiting for your certificate. I need to find time to generate them because I have a vision for how they should look. (23:45)

Alexey: The AI course and the new community I am working on should go together. The certificate should be a part of this new community. (24:08)

Revathy: You can design it as per your idea. (24:19)

Alexey: Tell us more about that course. (24:26)

Revathy: I became curious when I started following you and saw messages about an email course. I was wondering what an email course is compared to online or offline courses. Again, it was curiosity. One person from the group replied that every day you get an email about it. (24:32)

Revathy: Curiosity led me to enroll in this too. It was about AI Hero. Since I had already started on AI Dev Tools, let this be a continuation. It was also for free, so I enrolled immediately. (24:57)

Revathy: In that course, I developed a Q&A assistant which is able to query whatever questions you have about a Git repository. (25:24)

Alexey: You developed some AI agents? (25:30)

Revathy: Yes, it was a very useful thing. I was able to learn about chunking, how you get a document, and how documents are fetched from a repository and stored. I learned about querying, simple text search, and vector search. I learned how to do a semantic search. (25:30)

Revathy: There are concepts like cosine similarity that go deeper and deeper. To satisfy my curiosity, I would go and check the math behind cosine similarity and how word similarity works. I completed a project in that too. That was a very easy course that anyone could do. (26:03)

Revathy: Your materials were always very easy, Alexey. That is the biggest plus. Even for someone who does not know, they just have to follow it step by step. I think even an eighth grade student would be able to do it. (26:33)

Alexey: Thank you for the nice words. I am not sure about an eighth grade student; I would have to check that. Nobody who was a ninth grade student has approached me yet. Let me know if your son takes it. (26:53)

Alexey: You now work as an AI engineer. AI Hero sounds like it was the most relevant course for what you do. Can you tell us more about what you do right now and how you managed to get this job? Half a year ago you were wondering what to do, and now you have a job. (27:16)

Alexey: How did you understand this was the kind of job you wanted? How did you end up doing what you do and what was required? (27:42)

The Interview Process: Navigating a 7 Year Career Break.

Revathy: In fact, I hardly remember when I applied for this job because every day I would apply for around ten to fifteen jobs on whatever websites appeared. Wherever I found AI or ML engineer, I would apply for it. One fine day, the HR from the company messaged me saying it is a startup and asked if I would be interested. (28:00)

Revathy: I said I was okay with it, and she asked me to share my GitHub profile. I shared it. They scheduled an interview after seeing my portfolio projects. There was an online first round interview where they wanted to understand my background and my long career gap. (28:26)

Revathy: They were checking on that, and I told them for family reasons I took a break. I was very honest in my decisions and my career break. He asked about my Python experience. I told him I only started learning Python for the machine learning course and had only six months of experience. (28:54)

Revathy: I was very frank about being at a beginner level. For me, if they want to give me a job, let them, or else I will be happy building projects. With the knowledge of these three courses, I was very confident that I can build amazing projects. I told him that. (29:13)

Revathy: He looked at my previous salary and said as it is a startup, they would not be able to pay so much. I agreed to that because I wanted some flexibility in my timing. I negotiated on those terms. Then he asked if I would be ready to come for a face to face interview the next day, and I said okay. (29:37)

Revathy: I pinged the community because I was skeptical about the reduced pay and whether it would affect my career path. But I had mixed feelings because my career gap and my experience were almost equal. I worked for nine years and my break was seven years. I thought it was high time I started getting back to the line. (30:05)

Revathy: I felt I had to start at this point. I got flexibility at work so I can leave before my kids come back from school and then continue my work. I was okay with that and went for a face to face interview. He asked me to show my projects, and I showed him the multiclass obesity prediction project. (30:34)

Revathy: He wanted me to run the project on my laptop. He was inquiring about the dataset, the way I chose it, where I pulled it from, and everything I did in the project. I answered him and showed the REST web service URL and the output. He was okay with it. (31:00)

Revathy: Then related to Python, he asked a couple of questions. One was on string manipulation where I had to append a string. The other thing was reversing all the vowels in a given string. Only the vowels should be reversed. (31:36)

Revathy: I just wrote pseudo code for that. I explained that I would take it character by character, compare them, and put them in a stack and then pop them out. (31:58)

Alexey: Interesting. So first I need to find all the vowels and remember their positions, then revert them and insert them back? (32:10)

Revathy: Yes. When you print your string, only the vowels should be reversed. For example in my name, Revathy, I go character by character, take all the vowels and put them in a stack. At the time of printing, I check if it is a vowel, and if so, I go to the stack and pop it out. (32:22)

Alexey: That is simpler than what I had in my mind. (32:48)

Revathy: I just wrote pseudo code and he was happy with that. Then he asked when I would be ready to join the company. Basically, my nine years of experience was very crucial for him. He asked what I did there, and I told him I was a product owner managing a team of three to four people. (32:55)

Alexey: He valued your experience. It was only two interviews, right? One online and then one in person. But your role is AI engineer. (33:25)

Revathy: My designation is AI engineer. (33:36)

Practical Interview Tasks: Building a PDF Q&A Assistant.

Alexey: From what you said, it was not really an AI engineer interview. They asked about your machine learning project. (33:45)

Revathy: My first take home assignment was on AI. I was asked to build a RAG system. I had to develop a Q&A assistant that would give answers about a PDF document. (33:47)

Revathy: I had to use appropriate chunking strategies. Evaluation was based on retrieval accuracy and efficiency. This AI email course helped me for that. I just reused the project. (34:06)

Revathy: Where I had to use the GitHub repo, I used a PDF reader instead. I took the document and it was very easy. The next day, I was able to show my assignment to him. (34:30)

Alexey: That is really cool. I am happy to hear that the courses we do are so useful to people. So the process was online, then on site, then a home assignment, and then they asked when you could join? (34:43)

Revathy: Yes, that was it. (35:09)

Alexey: So what do you do now? (35:16)

Revathy: Currently, I am working on the LangChain framework and on RAG. It is related to the business domain. We have a lot of challenges in using AI for business purposes because AI can hallucinate. You have to have guardrails to get the best output for the business purpose. (35:16)

Alexey: Can you tell us a bit more specifically in which domain you work or what kind of problems you are solving? It is okay if you cannot be concrete. (35:47)

Revathy: I am not sure if I might run into trouble, but it is the healthcare domain. I joined on January 29, so it has been about three weeks since I joined. (36:04)

Alexey: Can you tell us what you learned in your job over this time? (36:22)

Revathy: I learned about ChromaDB. It was more like a continuation of my AI Hero course. I learned how to use RAG better and improved my chunking. I learned how to store chunks in a database and retrieve them for results. (36:27)

Revathy: The accuracy of results may not be as good as required for the business use case. The answer is not deterministic in some places, so you have to think of a better solution. It is all like a POC that is happening. I am in a learning phase and every day I learn. (36:52)

Revathy: I learn core Python skills because I have only six months of experience in Python. It is quite interesting and also challenging. (37:14)

Alexey: I am curious if you needed to do this again half a year ago, what would you optimize? Imagine you could time travel six months back and give yourself a note with some advice. (37:32)

Revathy: Six months back, what would I do? I think this has been the best journey. DataTalks.Club was the best part of my journey. I do not want to time travel back because the last six months were absolutely very fruitful for me. (38:21)

Revathy: It has built my confidence and helped me progress in my career. There is nothing I would want to go back and change. (38:46)

Alexey: Do you feel that you are behind with all the advances and the break you had? (38:59)

Revathy: Yes, I am very much behind. But with these courses now, I am on par with others. I feel this because I am with the latest trends in AI and machine learning. I am on the latest technology whereas I feel my colleagues are still working on old technologies. (39:06)

Revathy: Career Advice: Clear Plans, AI Mentors, and Hard Work (39:06)

Revathy: DataTalks.Club helped me a lot. (39:40)

Alexey: I am happy to hear this. Let us say you are not doing any time traveling, but what would you recommend to people who are listening? If they want to get an AI engineer role in three to six months, how should they go about that? (39:48)

Revathy: First, they should be clear on what they want. Once they know, they can look for courses. I would certainly recommend DTC. The courses are very structured and clear, even for a beginner. (40:06)

Revathy: I would also recommend them watching the videos of the podcasts that are available. There was a video about a full time mom who became head of data and cloud, Zia Ben. I listened to that podcast and was motivated by it. If she was in the same phase and able to do it, then Revathy, certainly you can also do it. (40:32)

Revathy: See more of such motivating videos and have a clear plan. You can make it yourself or ask AI to do it. Once you start the course, you have to devote a lot and put your heart and soul into it. You have to do hard work. (41:12)

Revathy: When you build amazing projects, I think you will definitely get your dream job. (41:29)

Alexey: Everything you said really resonates with me. First, you have to be clear on what you want. I often get questions about whether it is better to do machine learning or data engineering. The answer is that I do not know what is better for you; you have to decide. (41:42)

Alexey: Once you decide to do ML, I can help with how to do it. You have to decide yourself and try different things if you are not certain. Having some focus is helpful. You also said having a plan is good, and you can get one from ChatGPT. (42:04)

Alexey: I hear that from many different angles. When I was preparing for a marathon, the advice was that having a plan is way better than having no plan, no matter what the plan is. As long as there is structure and clear actions, you do not need to think. (42:51)

Revathy: I also referred to data science interviews. There was a GitHub repository which you shared that was very useful. There are also some coding questions there. In my free time, I used to go there and take a question, even a simple one like finding the biggest number from an array. (43:30)

Revathy: I start from the very basics. I used sites like HackerRank and CodeChef where there are beginner, intermediate, and expert levels. Don't waste your time and do something. (43:53)

Alexey: This was a part of the plan ChatGPT recommended to you. Learn this, do that, prepare for the interview, and do these coding assignments. At least you had the awareness of how to reach your goal. (44:13)

Closing Thoughts: Scaling the Learning Ladder.

Alexey: After we finish this interview, I will upload the transcript and ask ChatGPT to come up with a learning plan based on what we discussed. I am curious to see what will happen. Anything else you want to share before we wrap up? (44:30)

Revathy: I think we covered everything. Thank you so much for the courses, Alexey. You are doing an amazing job. I was not aware such great courses were happening for so long. (44:52)

Revathy: If I had known this before, maybe I would have started several years earlier in my free time. It is very hard to find a course where you understand things from the basics. Often, videos start from their own level of expertise rather than the basics. Machine learning was a course I enjoyed very much. (45:12)

Revathy: It progressed through AI Dev Tools and the email course. It was like climbing a ladder step by step for me. (45:37)

Alexey: I am really happy to hear this because I put a lot of effort into thinking about the sequence and how to make it accessible. My target audience was engineers who have no idea about machine learning. Preparing the best material for them gives me validation for the work I did. (45:49)

Revathy: Hats off to you and your team, Alexey. You are doing a brilliant job. The team is also very important. Whenever I speak with my friends, I mention these courses. (46:26)

Revathy: I tell them that even if they are just interested in knowing about machine learning or AI, they should subscribe for these courses. You deserve it. (46:32)

Alexey: It has been a pleasure talking to you and seeing you in the community. I hope that with work you will still find time to be around. It is really nice to have community members like you. I want to thank you for being active since September and for this interview. (46:45)

Revathy: Thanks for inviting me. (47:08)

Alexey: It was a pleasure talking to you. I guess we will see each other around. Thanks everyone for listening. Today was different since we typically have questions from the audience, but I still enjoyed it a lot. (47:14)

Alexey: I hope everyone listening will enjoy it too. Thank you. (47:27)

Revathy: Thank you. (47:32)