My First Job Search, 2025-26

Introduction

I spent about seven months on my first job search, from early August 2025 to early March 2026. I deliberately took a fairly ambitious route from the start. I applied to a lot of overseas roles and targeted companies with very high hiring bars. There were plenty of mistakes along the way, but I also learned a lot. In total, I applied to 97 positions and ended up with one offer. I wrote this post to share my thoughts and experiences from the whole process.

How I Chose Where to Apply

When deciding where to apply, I had three things in mind.

• Companies with high talent density: places where developers and nearby technical roles are filled with very strong people
• Companies with strong momentum: companies that were growing quickly, or belonged to industries that were growing quickly
• Companies with strong technical depth: companies with top-level technical strength in a particular area of computer science

By those standards, quant was one of the best fits for me. The industry tends to have both high talent density and a high technical bar. I had also interned at a small hedge fund before, so it felt more familiar than most other areas. Outside of quant, the two fields I was most interested in were AI and blockchain.¹ I thought both had strong growth potential.

I also applied aggressively to overseas roles. Since talent density mattered a lot to me, I wanted to work at companies that pulled together strong people from different countries.

During the Job Search

Early Mistakes

Early on, roughly from August through October, I did not really understand what interviews were actually going to test. I was reasonably prepared for algorithms and data structures, but I was far less prepared for core CS topics, especially operating systems, networking, and C++. I was also underprepared for system design.

In general, I prefer jumping into things early even if I am not fully ready. I tend to learn faster by running into real problems than by preparing alone forever. In the long run, I still think that approach often works better. But in this job search, I think it was the wrong call. One of my bigger regrets is that I applied to the very best companies I could go after before I was actually ready for them.

Once an interview got scheduled, I had to cover several topics in one or two weeks. That kind of rushed knowledge was shallow and fragile, and I think that showed in the interviews.

Still, that period gave me something too. Before this process, I did not have a very clear picture of what I wanted to do after graduation. During this process, at least the direction became much clearer. In particular, that was when I became much more certain that I wanted to be a systems engineer.

What I Studied

Until pretty recently, I did not know much beyond algorithms and data structures. So during the job search, I spent a lot of time trying to fill in the gaps.

The first thing I read was Algorithms for Modern Hardware. It explains how to write efficient code while taking CPU behavior and memory structure into account. I think it is especially good for someone who is strong at algorithms but still weak on systems knowledge. Then I read Operating Systems: Three Easy Pieces. Because I was under time pressure, I did not read all of it. I focused on the chapters that seemed most likely to come up in interviews. Later in the process, I also studied system design with System Design Interview. Again, I did not read every chapter. I just picked the ones that seemed most important.

I also watched a few C++ conference talks. For example, I watched this one. For hands-on practice, I used highload.fun quite a lot. It gives you tasks where the goal is to write code that finishes computation-heavy work as fast as possible. Out of everything I did during the process, that was probably the most fun.

I also used LLMs a lot. I had them keep generating questions about core CS topics, and I answered them one by one. I asked them to adjust the difficulty based on my answers and break concepts into chunks small enough for me to reason about. That turned out to be pretty helpful for interview prep.

Looking back, I think the whole idea of cramming systems knowledge for “interview prep” was wrong. What I actually needed was not interview prep in that narrow sense, but a real understanding of systems.

What I Got Wrong in Interviews

I already knew the usual advice from the beginning: think out loud, write readable code, and so on. I think I did fairly well on those basics.

But after a number of interviews, I realized there were a couple of problems in how I approached them.

One was that my answers were too passive. I had a tendency to answer too briefly and too carefully. Part of it was a vague assumption that if the interviewer wanted more, they would ask. Another part was that when I felt unsure about a topic, I would instinctively retreat and try to stay safe. Toward the end of the job search, I tried to fix that consciously.

Another issue was that I did not give myself enough time to think. I felt like I had to keep talking, and because of that I sometimes answered too quickly even on questions that were clearly testing reasoning rather than recall. I think this showed up especially in some quant interviews. I tried to fix this too toward the latter half of the process.

Miscellaneous

• Whenever I got an interview, I tried to research the company as much as possible and understand its business. This was especially true for smaller companies.
• I had already been practicing English conversation for a few years, but English interviews were still harder than Korean ones. Honestly, even “How are you?” still feels awkward.
• I sometimes tried to think from the recruiter’s side. It did not help much.
• Getting a job is not a Codeforces rating contest, not even in quant.
• I sometimes wondered whether understanding core subjects like operating systems, networking, and data structures/algorithms will still matter as AI keeps advancing. I still do not know the answer.
• From the company’s point of view, hiring a new grad seems expensive and risky.
• Thinking too hard about interview attitude is probably not that useful. It seems better to focus on actual ability instead.

Main Results

Results

I applied to 97 positions, interviewed with 10 companies, and got one offer.

Company	Location	Outcome	Role
Moloco	Seoul 🇰🇷	Round 1, rejected 🟥	SWE Intern
Toss	Seoul 🇰🇷	Round 1, rejected 🟥	Server Developer
Jane Street	Hong Kong 🇭🇰	Round 1, rejected 🟥	SWE
HRT	Singapore 🇸🇬	Round 2, rejected 🟥	SWE (C++)
Jump Trading	Singapore 🇸🇬	Round 2, rejected 🟥	C++ SWE Intern
think-cell	Berlin 🇩🇪	Round 2, rejected 🟥	C++ Developer
Presto Labs	Seoul 🇰🇷	Round 1, rejected 🟥	Quant Developer
SNJ LAB	Seoul 🇰🇷	Round 2, rejected 🟥	System Trading Developer
Headlands	Amsterdam 🇳🇱	Round 2, rejected 🟥	Software Developer
FuriosaAI	Seoul 🇰🇷	Final round, accepted 🟩	SWE (Inference Engine)

Applications by Category

I applied to 27 tech companies and 46 quant trading firms. I also sent my resume to 5 headhunters. Most of my applications were for general SWE roles, and 12 were for quant developer/researcher/trader positions.

Overseas Applications

84 of my 97 applications were for roles outside Korea.

My interview rate was highest with Korean companies. Among overseas applications, I got the most interview opportunities in Singapore and Hong Kong. I also got interview opportunities in the Netherlands and Germany, but none in the UK.

I applied to 23 US positions but did not get any interviews. I thought it was very unlikely that a company would pay something like $100,000 to hire a new-grad software engineer, but I applied anyway because I figured I had little to lose.²

After the Offer

FuriosaAI had been on my radar for years. If I was going to start my career in Korea, it was one of the companies I wanted most. I was very happy to get an offer from a company like that, and I ended up joining.

Since getting the offer, I have been trying to understand the AI accelerator industry more broadly. Over the past few years, AI data centers have been growing very quickly. And it is not just frontier AI labs anymore. Demand from many companies, governments, and public institutions to build their own AI infrastructure is growing too. From that angle, the long-term outlook for the AI accelerator industry looks pretty strong to me. Korea also feels especially interesting here. It is one of the few countries trying to build AI infrastructure with its own chips, and that makes me think the domestic AI accelerator industry has strong momentum.

After I join, I want to learn as much as I can and do my job well. At the same time, I do have some honest worries. There are clearly a lot of very strong people at FuriosaAI, and I do wonder whether I will be able to pull my own weight among them.

More broadly, when I think about the performance of recent LLM coding agents and the speed at which they are improving, I find myself thinking a lot about the future of software engineering. Even if I keep the same title, the actual work may end up looking very different from what it looks like now. That feels exciting and scary at the same time. For the next few years, I want to stay sharp and do my best to ride that rapid change well.