It was late, but starting from the first semester of senior year I began shaping my career direction. With the mindset of “let’s do something, anything,” I spent a breathless period during the semester: teaching practice, writing a paper on “using AI to improve every student’s performance in information-subject classes,” building an information-subject exam system, and participating in the SNU competition at the same time.

Perhaps because of that, I did not achieve strong results such as winning a competition or winning a paper contest. Afterward, over roughly four months, I participated in hackathons and campus hackathon competitions, and I was at competition venues once or twice a month. In some of those competitions, I achieved good results.

In the past, I thought hackathons were extremely difficult competitions only for programming monsters, so I kept avoiding them. But after actually participating, I realized that some competitions are like that and some are not. Only in the second semester of senior year did I understand that I could choose competitions that fit my current skill level and try to produce results there.

Going through this breathless period made me think about many things.

Mixing with people I did not know, sharing the common goal of winning an award, going through trial and error, and somehow producing a result is an important experience. As a student, it was a rare opportunity to gain valuable experience for free without too much pressure, so I tried to achieve the goal as much as possible.

However, most of the actual development output came from vibe coding. It was difficult to build something in a short period like one night and two days or two nights and three days while collaborating with strangers, and even code management through GitHub was almost impossible.

Also, the teams that won hackathons were usually the teams with good ideas. I felt that the important factors were whether the implementation was defined concretely and technically, and how well the team communicated that through PPT and presentation skills.

As I wondered how to improve communication and presentation skills, I also realized that short-cycle hackathons would no longer help me very much. Of course, this is only my personal opinion.

I still think receiving an “award” is important. It adds one more story to talk about in a resume, portfolio, or interview. And the rhythm of building an idea through a short team sprint and carrying it through to a presentation was very important. But from the perspective of code and engineering, there was not much I could talk about. So if I join another competition later, I want it to be a longer-term one.

After the second semester ended and vacation began, I started working on the projects I want to build while strengthening my technical expertise.

---

It is clear that the number of developers will decrease. As a student who wants to become an AI Product Engineer, I want to build something that becomes a real end-to-end service, which I have not yet done.

One project idea I have is based on this thought: “If the internet is cut off someday, if the internet is overtaken by AI, or if a natural disaster takes away communication infrastructure, then we will need something that allows communication without the internet and contact without base stations.”

From that idea, I want to build a service called Mesh AI Community, a decentralized intelligent communication solution that works even in extreme environments where centralized infrastructure is absent or networks are disconnected. I know the difficulty is high, and because I lack domain knowledge in this field, at the moment I can only plan the project at a surface level. However, if I join this program, I expect to receive help from students and mentors with diverse backgrounds, so I want to write and develop this project idea. I have also started studying signal processing from the moment I decided to apply, and if I join the program, I expect to learn it faster through study sessions.

What I want to build is a service whose core technology combines on-device AI, such as Gemma 3n or an Ollama-based lightweight LLM, with a P2P mesh network using BLE or LoRa direct communication. It should enable more private communication without internet intervention and communication in any environment. When users face difficulty, AI would support real-time summarization, translation, and location sharing. Based on the Shannon-Hartley theorem, the service would improve communication stability through channel-capacity optimization, such as efficient modulation under low SNR. Its strengths would include information sharing among survivors during disasters and privacy protection through local processing.

Below is the plan for achieving this goal.

1. Preparation phase: Form a team with a network/signal-processing trainee from an electrical/electronics background and a CV/ML engineer. Build a LoRa SDK/Raspberry Pi testbed, develop a BLE mesh prototype as a mini project, and study Shannon-Hartley application through mentoring.
2. Main phase (3 months): Implement the app with Rust/Flutter, including messaging and AI edge inference. Run simulation tests with a target of 1 km range and 80% packet success rate.
3. Evaluation/deepening: Deploy a demo service, test it in remote areas, and complete the result as a portfolio project.