Seokho Lee

Overview

NextGen is an academy platform for developing the next generation of creators. It manages web, API, and mobile applications in one monorepo and includes the production architecture required for payments, video, email, SMS, file uploads, and deployment.

The course platform had to launch as a revenue-generating production service within about one month while meeting DRM requirements. I was responsible for the implementation and operational readiness needed to launch it on schedule.

The documented architecture includes a Next.js web application, Fastify API, Expo mobile app, PostgreSQL and Redis, S3 and CloudFront, Mux, TossPayments, Resend, Naver SENS, Terraform, GitHub Actions, Vercel, and AWS EC2/ECR deployment.

My Role

• Organized the Next.js, Fastify, and Expo monorepo architecture and development environment.
• Built a Turborepo workflow connecting the API, web, mobile, and shared packages.
• Designed and implemented the core product flow connecting course access, video protection, payment, user authentication, and production deployment.
• Connected AWS EC2/ECR, nginx Blue-Green deployment, Vercel prebuilt deployment, Terraform infrastructure, and GitHub Actions CI/CD.
• Documented development and production environment variables, SSM Parameter Store, Secrets Manager, RDS, and S3/CloudFront synchronization.

Quantified Results

• Completed the course platform to a launch-ready standard within approximately one month of development.
• Recorded a 0% error rate at peak concurrency of roughly 1,000 users.
• Developed and operated a service with up to approximately KRW 60 million in daily revenue.
• Organized three applications—web, API, and mobile—along with five shared packages for types, utilities, API clients, authentication, and configuration.
• Designed GitHub Actions workflows for CI, development deployment, production deployment, iOS TestFlight, and iOS App Store candidate releases.
• Consolidated more than six operational layers—including AWS, Vercel, Cloudflare, EAS, SSM, and Secrets Manager—into one deployment guide.

Overview

BL Agent is a desktop application for freight-forwarding operations that extracts key information from B/L PDF documents, supports human review, and saves the results into an existing .xls template. A React interface and Tauri/Rust backend package document intake, extraction status, review, and XLS export into one operational tool.

Rather than building a simple OCR demonstration, I focused on deliverable production quality, including model configuration, extraction status, cancellation and restart behavior, and Windows installer packaging.

My Role

• Rebranded the existing BL Reader product as BL Agent and aligned application metadata, release artifacts, documentation, and persisted storage keys.
• Designed the PDF-processing, field-review, result-storage, and XLS-export workflows.
• Directly resolved desktop distribution issues involving Tauri v2 permissions, Windows NSIS installers, and macOS application bundles.
• Wrote clear Korean operational copy and separated workflow states such as processing, review, and save so users could understand the tool immediately.

Quantified Results

• The application currently performs work equivalent to the workload previously handled by six employees at the client company.
• Verified macOS application bundles and Windows installer artifacts across the 1.0.x release line.
• Operated a validation workflow covering both the React/TypeScript frontend and Rust/Tauri backend.
• Resolved multiple release-blocking issues involving SQLite identifier policy, extraction-state synchronization, window-close permissions, and NSIS packaging.

Overview

WHOFA is a service connecting fans and artists. I developed new profile and community features for its iOS app in Swift, maintained existing functionality, and later led the migration to Flutter. The product includes authentication, community, Preplay, profiles, rewards, notifications, and media processing.

At WHOFA, I owned work from iOS feature development through the Flutter migration, while treating navigation, media interaction, and community usability as product-quality concerns.

My Role

• Developed new profile and community features in Swift and maintained the existing iOS application.
• Led the existing iOS application's migration to Flutter and Riverpod, implementing its screen and state architecture.
• Worked across Preplay detail, iOS-style navigation, deep linking, on-device databases, and media processing.
• Resolved user-facing issues including image downloads, audio attachment crashes, relative-time errors, tab selection defects, and stale state initialization.
• Improved platform-specific UI details including iOS blur, refresh indicators, and Hero animations.
• Brought features to a user-visible level of polish, addressing crashes, navigation, animation, and media edge cases rather than stopping at basic feature completion.

Quantified Results

• Led the existing iOS application's migration to Flutter within a four-person product team.
• Implemented new profile and community features in Swift and maintained the existing iOS application.
• Handled multiple operational user touchpoints—including authentication, community, media, notifications, and rewards—within one production product.

Overview

Interactor is a product for building automated workflows by connecting Connectors for multiple SaaS services. While interning at the Silicon Valley company Pulzze Systems, I developed the Connector UI frontend and improved its test screens and data-type description files.

During the internship, I worked for two months in Sunnyvale and three months at the Seoul office.

My Role

• Improved frequently used interfaces including Connector selection, Action and Authentication tests, developer navigation, and Billing screens.
• Addressed cases where test results did not refresh and internal errors were not clearly exposed, improving the debugging experience.
• Authored and reorganized data-type description files for Connector families including Google, Slack, Salesforce, Notion, HubSpot, Jira, Mailchimp, and Microsoft.
• Supported the Connector development experience through work such as adding a FormData dependency and improving README documentation.
• Focused on reducing friction in testing and selection flows so both Connector developers and users operating the automation product could navigate the system more confidently.

Quantified Results

• Worked with data-type descriptions across more than eight service families, including Google, Slack, Salesforce, Notion, HubSpot, Jira, Mailchimp, and Microsoft.
• Improved at least four core administration surfaces: Authentication tests, Action tests, Connector selection, and Billing.
• Experienced the development workflow of a production SaaS product over a total of five months in Sunnyvale and Seoul.

Overview

Toongether is a web and mobile platform that lets webtoon creators publish freely without complex contract procedures. It aimed to provide creators with publishing opportunities and an automated release environment, while giving readers convenient discovery and reading experiences across a broad range of genres.

The team consisted of one iOS developer, one Android developer, two web developers, and two server developers.

My Role

• Developed and released the iOS application.
• Created FlowKit, a push/pop-based navigation library designed for use with TCA.
• Led product direction, planning, design, development, deployment, and marketing as team leader and PM.
• Conducted tests and interviews with webtoon creators, the product's primary users, and incorporated their feedback into the product.

Quantified Results

• Received the Top Excellence Award at the SK Smarteen App+ Challenge 2023.
• Incorporated feedback from approximately 300 readers and around ten creators.
• Served as leader and PM for a six-person team spanning iOS, Android, web, and server development.
• Extracted FlowKit into a standalone Swift Package so it could be reused beyond Toongether.

Overview

DodamDodam is a smart-school service that makes daily life more convenient for students and teachers at Daegu Software Meister High School. It moved offline administrative and student-life processes onto the web and mobile apps, including meal information, leave and overnight requests, merit and demerit points, wake-up-song requests, and late-night self-study.

The team consisted of five iOS developers, five Android developers, six web developers, and four server developers.

My Role

• Developed and released the DodamDodam iOS application.
• Maintained the app for approximately one year and shipped ten versions, including late-night self-study and a meal-information widget.
• Collaborated with the Android Chapter to align the design of the two mobile applications.
• Led project management, meetings, documentation, and coordination across development chapters as PM.

Quantified Results

• Contributed to a service used every day by approximately 200 students and 34 staff members.
• Released ten iOS versions during one year of maintenance.
• Served as PM in a multidisciplinary team of five iOS, five Android, six web, and four server developers.
• Chose to publish source code in order to support the development skills of other students at the school.

Overview

Love & Code is a visual novel game that presents the first collaborative project experience of new developers in an approachable, entertaining form. Drawing on the real experiences of Software Meister High School students and aspiring developers, it combines a realistic story with AI chats built around each character's persona.

The team consisted of three developers and two illustrators.

My Role

• Led the migration to Flutter so Android, iOS, and desktop versions could be developed from one codebase.
• Designed the custom Story format and library for expressing visual-novel scenarios.
• Designed and developed character-persona AI chat using GPT-4o mini and Ktor.
• Led team management, planning, design, development, beta testing, and marketing as PM.

Quantified Results

• Received the Excellence Award at the SK Smarteen App+ Challenge 2024.
• Simultaneously served as PM and a core client developer in a team of three developers and two illustrators.
• Consolidated the initial KMP and Compose architecture into a Flutter codebase spanning Android, iOS, and desktop.
• Implemented interaction beyond a conventional visual novel through the Story format and AI chat capability.

Overview

Allermi helps users identify risks while choosing food by comparing products with their allergy information. The goal was to make food ingredients and allergens understandable even to people who were not already familiar with allergy terminology.

The Allermi Project organization contains separate iOS, Android, and Server repositories, and the project received the Mutual Growth Award at the Samsung Junior Software Creation Contest 2022.

My Role

• Contributed to iOS development, implementing the flow for users to review and manage allergy information.
• Designed mobile screen structures and information presentation for quick risk assessment.
• Coordinated API and interface requirements with the server and Android teams, gaining experience with a multi-platform team architecture.

Quantified Results

• Received the Mutual Growth Award at the Samsung Junior Software Creation Contest 2022.
• Completed a team project divided across three repositories: iOS, Android, and Server.
• According to Samsung SDS's public competition materials, the team was selected as one of the award winners from 60 finalists in the 2022 competition.

Overview

Desktop Fushi is a personal project that implements the character FUSHI as a desktop pet that moves naturally across the user's workspace. Instead of placing the character inside a conventional application window, the project uses platform-specific graphics APIs and window layering to integrate FUSHI with the desktop itself.

My Role

• Designed a desktop overlay architecture that accounts for both Windows and macOS.
• Prototyped Direct2D rendering on Windows and Metal-based rendering on macOS.
• Investigated desktop-pet-specific interactions including transparent windows, click-through behavior, animation-frame rendering, and screen-boundary handling.

Quantified Results

• Created an initial multi-backend architecture that accounts for rendering paths on both Windows and macOS.
• Organized Direct2D and Metal integration within a single project architecture.
• The project remains in active development, with platform-level technical validation prioritized over final performance measurements.

Overview

Windows 11 for Galaxy Tab 6 Lite is a porting project intended to run Windows 11 on ARM on an Exynos 9611-based Galaxy Tab S6 Lite. It is a low-level systems project that analyzes the Android tablet's boot chain, ACPI tables, drivers, and device-tree-like hardware information, then reshapes them into forms Windows can recognize.

My Role

• Analyzed the target device's hardware configuration and boot flow.
• Combined ACPI ASL, ARM64 assembly, C, and Python tooling to conduct the experiments required for a Windows on ARM port.
• Built a repeatable validation workflow using ADB and a custom recovery environment to test images and configuration changes.

Quantified Results

• Conducted low-level experiments to validate the feasibility of Windows 11 on ARM on the unconventional target of an Exynos 9611-based ARM64 tablet.
• Documented a porting workflow that combines C, ARM64 Assembly, ACPI ASL, Python, WDK, and ADB.
• The project remains in progress, prioritizing step-by-step validation of boot and device recognition over a completed feature set.

Overview

App Pilot is a no-code application development service that lets users combine components through touch and drag interactions to create a real iOS app. It was selected as a winner of the Apple WWDC Swift Student Challenge 2024 and was built as a WYSIWYG editor that runs inside Swift Playgrounds.

Rather than stopping at visual screen composition, the editor represents the user's layout through concepts aligned with SwiftUI code and exports it back into a runnable application.

Following Allergist, this became my second Apple Swift Student Challenge win and led to an invitation to Apple Park during WWDC24.

My Role

• Designed the entire application architecture and editor experience, then implemented it in SwiftUI.
• Created proxy classes for major UI components such as Image, VStack, and Text to render user-authored work.
• Analyzed the .swiftpm structure that Xcode can open and built an export feature that emits a user's application in the same format.
• Connected preview and code generation into a complete experience in which the screen a user builds becomes a real app.

Quantified Results

• Selected as a winner of the WWDC Swift Student Challenge 2024.
• Achieved two consecutive Apple Swift Student Challenge wins.
• Connected editing, preview, and code export in a single Swift Playgrounds project.
• Reduced the cost of adding new components by separating rendering and export logic through a component proxy architecture.

Overview

Allergist is an iOS service that lets users check food allergens against their personal allergy information. It grew out of my own experience living with food allergies and was selected as a winner of the Apple WWDC Swift Student Challenge 2023.

To fit the U.S. competition context, I used U.S. allergen classifications and wrote every user-facing string in English.

My Role

• Designed the service concept, data model, search experience, and settings interface, then implemented the complete product in SwiftUI.
• Built food-list exploration around each user's selected allergens so that risky items could be identified immediately.
• Bundled the food dataset inside the app so the experience would work in the offline Swift Playgrounds environment.
• Optimized the loading pipeline so the full dataset remained searchable within the competition's application size limit.

Quantified Results

• Processed 558,728 food records entirely within the application.
• Reduced initial data loading from approximately one minute to approximately three seconds.
• Selected as a winner of the WWDC Swift Student Challenge 2023.
• Became the first project in a two-year run of Apple Swift Student Challenge wins that continued with App Pilot.

Overview

FlowKit is a navigation library that brings explicit operations such as push, pop, popToRoot, replace, sheet, and alert to SwiftUI in a way that resembles UIKit's navigation stack. It began as a solution for reliable navigation in the complex Toongether application and was later packaged as a reusable Swift Package.

My Role

• Designed an API that bridges SwiftUI's declarative view model with imperative navigation requirements.
• Combined SwiftUI with UINavigationController to support iOS 13 and later.
• Added dependency-based usage examples so navigation could be triggered from reducers in TCA (The Composable Architecture).
• Prepared the README, installation instructions, and examples so the implementation could be used as a standalone package rather than an internal project utility.

Quantified Results

• Packaged the library as a Swift Package supporting iOS 13 and later.
• Unified more than seven core navigation operations—including push, pop, popToRoot, replace, reload, sheet, and alert—behind one API.
• Adopted by at least three school projects.

Overview

Story is a library for representing and processing visual-novel direction—including dialogue, backgrounds, effects, branches, and scene progression—through an XML-based format. I created it for Love & Code to support custom story presentation and effects, then expanded the workflow with a human-friendly StoryMaker GUI editor.

My Role

• Defined Story's XML format around the scene data and commands required by a visual novel.
• Built a library that parses XML files into structures the game runtime can execute.
• Implemented the StoryMaker desktop GUI editor in Flutter so writers and developers would not need to edit XML by hand.
• Designed the format to connect with the gameplay, special effects, and persona-conversation flow in Love & Code.

Quantified Results

• Designed the format for actual story processing in Love & Code.
• Validated the format, parser, and editing workflow together by delivering both a library and a GUI editor.
• Improved game-development productivity by carrying even small supporting features through to a complete, usable state.

Overview

Amaze is a multiplatform game engine built in Rust. Its documented scope includes in-house rendering and audio HALs, physics, networking, text shaping, ICU internationalization, OpenXR, an MCP server, and a shader compiler.

The goal is to support macOS, iOS, Windows, Linux, Android, and the web from a single source tree, while treating MCP and SKILL bundles as first-class capabilities so AI agents can work directly with the engine.

My Role

• Defined the project's technical direction and workspace architecture.
• Separated engine, platform, backend, and foundation crates to create an extensible structure for long-term development.
• Clearly documented the intended scope of rendering, audio, physics, XR, networking, and MCP tooling, along with the distinction between implemented scaffolding and planned functionality.
• Designed a developer experience centered on documentation, examples, and skill bundles that AI agents can understand and use.

Quantified Results

• Defined a target platform matrix covering six operating-system runtimes, five GPU backends, and five audio backends.
• Designed a workspace containing 22 examples and a skill bundle for AI agents.
• Decomposed major engine capabilities—including determinism, XR, world streaming, networking, and MCP tooling—into crates that can be implemented and maintained over the long term.