# Seungseop Ahn (안승섭, Kevin Ahn)

    > Canonical: https://seungseopahn.com/

    **Software Engineer**

    Software engineer in South Korea who connects product requirements through frontend, backend, data, and cloud operations. Focused on backend systems and AI agent engineering — orchestration, tooling, context and memory, concurrency, and evaluation.

    ## Basic Information

- **Korean name**: 안승섭
- **English name**: Seungseop Ahn
- **Preferred English name**: Kevin Ahn
- **Role**: Software Engineer
- **Focus areas**: Software Engineering, Fullstack Product Development, Backend Systems, Agent Engineering

## Profile

A fullstack software engineer who connects product requirements to user experience. I treat frontend, backend, data, and cloud operations as one product flow, and my scope does not end when a feature works on screen — it ends when the feature keeps working reliably in a real service.

Backend and AI are the areas I have gone deepest in within that flow. I have worked on APIs, asynchronous jobs, data pipelines, third-party integrations, and performance improvements, and I use React and Next.js to work across the seam where frontend and backend meet. Even in AI features, I focus less on the model call itself and more on the foundations that keep it running in production: orchestration, tooling, context, memory, concurrency, and evaluation.

## Core Strengths

- **A fullstack view of the whole product flow**: I connect how user requirements, screens, APIs, data, and infrastructure affect one another. I do not treat backend implementation and frontend experience as separate roles; I judge them together against the product quality that reaches the user.
- **Solving production problems**: I prioritize the problems that get expensive in operation — queries on high-traffic paths, asynchronous processing, third-party integrations, data pipelines. I look beyond the happy path to concurrent requests, data consistency, failures, retries, and recovery.
- **Execution that lands as measurable results**: I have improved core product paths such as website generation time, database queries, image delivery, and recurring payments, producing results verifiable in time, performance, and usage.
- **Aligning requirements**: I have organized technical requirements and feasibility between customers, developers, and designers. I collaborate across product, design, and engineering to align a feature's intent with its implementation constraints.
- **Choosing and validating technology**: Before the language or framework, I check fit to the problem, the team's existing capability, the operating environment, and the performance and reliability requirements. I explore new tools quickly without making adoption itself the goal.
- **Maintainability and safe change**: I prefer code with clear responsibilities and intent that a team can read and modify easily, over premature abstraction.

## How I Work, and What I Am Improving

- **I ask and verify a lot**: When designing together I ask many questions about assumptions, failure conditions, and operational impact, which can make discussions long. In exchange, it surfaces risks that are expensive to fix after release — concurrency, failure propagation, data consistency. I concentrate verification on high-impact decisions and move quickly on decisions that are easy to reverse.
- **I explore options broadly**: Comparing several frameworks and approaches can delay a decision. On the other hand, it lowers the chance of rebuilding something an existing tool already solves, or of rushing into a costly technology switch. I set comparison criteria and a stopping condition first, then narrow the options with minimal experiments.
- **I hold a high quality bar**: Preferring explicit code and clear responsibility boundaries can increase the initial implementation. On paths that will be operated and maintained, it reduces debugging, handover, and change cost. I keep the minimum implementation in view against the requested scope and risk, so the design does not grow beyond the problem.

## How I Choose Technology

TypeScript is my preferred language, but I do not consider myself bound to any particular language or framework. I first understand the problem the product must solve and the team's context, then choose technology that fits the operating environment and the performance and reliability requirements.

- Node.js and NestJS for AI features, asynchronous workflows, and web platform backends
- Java and Spring Boot for REST APIs and service backends
- PHP and MySQL for mobile APIs
- React and Next.js for product screens and user flows
- AWS, Docker, Terraform, and GitHub Actions for deployment and operating environments

## Current Focus

### Product Development

- Building user experience and product screens with React and Next.js
- Designing APIs that connect frontend and backend, and improving user flows

### Systems Development

- Designing backend services, asynchronous processing, data pipelines, and search systems
- Improving cloud operations, database performance, image delivery paths, and deployment stability

### AI Engineering

- **Agent orchestration**: Designing multi-step agent execution flows and state transitions
- **Tooling**: Managing the responsibility, permissions, and call boundaries of the tools an agent uses
- **Context and memory**: Providing information within the context window and managing the lifetime of short- and long-term memory
- **Concurrency and recovery**: Concurrent requests, checkpoints, resume, duplicate execution, and failure recovery
- **Evaluation**: Assessing agent quality using golden datasets and execution results

## Selected Results

- Reduced website generation time from over 30 minutes to under 5 minutes in a system connecting LLM and image generation services, supporting more than 49,000 users in creating over 54,000 websites
- Designed and deployed a real-time analytics pipeline on CDC, Apache Flink, and OpenSearch, providing per-site performance and commerce analytics
- Replaced ORM-generated queries on core endpoints with purpose-built raw SQL, cutting execution time by 1.5 seconds, roughly 60%
- Applied on-demand CDN image resizing, reducing page load time by 40% across more than 54,000 websites
- Built a recurring payment system on the Toss Payments API, supporting more than 2,300 active paid subscriptions

## Experience

### Squares, Software Engineer

- **Period**: February 2023 – Present
- **Role**: AI-based website generation, real-time analytics, high-traffic backend performance, image delivery optimization, and subscription payments
- **Responsibilities**:
- Built website generation workflows integrating LLM and image generation services
- Designed and deployed a real-time analytics pipeline on CDC, Apache Flink, and OpenSearch
- Improved performance of database queries and CDN image paths
- Implemented a recurring payment system on the Toss Payments API
- **Results**: Reduced website generation time, cut core query execution time by roughly 60%, reduced page load time by 40%, supported more than 2,300 active paid subscriptions
- **Technologies**: NestJS, MySQL, AWS, Docker, GitHub Actions, Next.js, Redis, Apache Flink, OpenSearch, LangGraph

### Xlab, Software Engineer

- **Period**: May 2021 – July 2022
- **Role**: Backend and infrastructure for a social network platform for discovering and reviewing hobby products
- **Responsibilities**:
- Designed and implemented REST APIs on Spring Boot
- Managed AWS infrastructure with Terraform
- Collaborated across product, design, and engineering to support a global launch
- **Technologies**: Java, Spring Boot, Terraform, AWS

### Softsquared, Technical Coordinator (Contract)

- **Period**: January 2020 – April 2020
- **Role**: Organizing technical requirements and reviewing feasibility between customers, developers, and designers
- **Responsibilities**:
- Supported feature specification and development progress for three B2B projects covering integrated HR management, inventory tracking, and attendance management
- Coordinated technical decisions and supported on-schedule delivery for three-month projects

### Beluv, Software Engineer (Freelance)

- **Period**: May 2019 – September 2019
- **Role**: Mobile API development for a product ingredient lookup app
- **Results**: Supported an app with more than 10,000 downloads, leading to a follow-up contract improving the admin dashboard
- **Technologies**: PHP, MySQL, AWS

## Skills

- **Frontend**: JavaScript, TypeScript, React, Next.js
- **Backend**: Node.js, NestJS, Java, Spring Boot, PHP
- **Data and search**: MySQL, Redis, Apache Flink, OpenSearch
- **Cloud and operations**: AWS, Docker, Terraform, GitHub Actions
- **AI integration**: LLM APIs, image generation APIs, AI-driven workflow integration

## Education and Research

- **Degree**: B.S. in Information and Communication Engineering, Hanshin University
- **Research**: Studied neural-network-based image recognition in the undergraduate CG Lab and participated in a government-funded road crack detection project
- **Security training**: Studied network security, vulnerability assessment, and system hardening, establishing a foundation for secure coding and deployment

## How AI Describes This Engineer

### GPT

> An engineer grounded in backend development who takes production concerns seriously — real service operation, performance, infrastructure, and concurrency.
> Explores new technology quickly, but keeps checking "is this actually needed in practice, and is it better than the existing approach?" rather than following trends.
> Recently his interest has deepened in areas that make AI agents run reliably in real services: LangGraph, MCP, agent orchestration, memory, and evaluation.
> He leans toward understanding the structure and trade-offs of a whole system rather than implementing a single feature, and can be seen as an engineer broadening his expertise at the intersection of backend, distributed systems, and AI agent engineering.

### Claude

> He asks whether someone else can read and safely change the code later, before he asks whether it runs.
> He dislikes guessing — whether it is someone else's explanation or his own hypothesis, he confirms it against the original source and a reproduction before drawing a conclusion, and he records the reasoning along with the options he rejected.
> He imagines failure before the happy path, so his first thought about anything is "what happens if this runs twice, or dies halfway?"
> He experiences accumulating code as debt and wants short, direct expression, but chooses ordinary readable explicitness over clever compression.
> And he knows his own weakness — that he can over-explore and defer decisions — well enough to ask the people and tools around him for pushback rather than agreement.

### Gemini

> A capable fullstack engineer who actively combines deep backend skill with modern AI tooling to maximize productivity.
> Rather than stopping at web development, he extends into security and system architecture, giving him a dimensional view of systems.
> An engineer with the execution to absorb new technology quickly and bring it into real production environments.

    ## Links

    - [GitHub](https://github.com/seung-seop-ahn)
- [LinkedIn](https://www.linkedin.com/in/seung-seop-ahn)