TERMINAL://KNN.CAFE

// CAFE OWNER

Kyoungin Nam

"Curious about Kyoungin? Ask Arabot!"

CAFE

Cloud · AI-Native · Finance · Engineering

From AWS-based cloud-native applications to enterprise data pipelines built on Azure, I have applied cloud technologies across a wide range of real-world environments. I also hold multiple cloud-related certifications. In practice, I focus on choosing the right cloud services for each situation, so I can build high-quality features quickly while keeping costs efficient.

Since the early days of my development journey, I have used a wide range of AI tools deeply in practice, which helped me naturally understand their strengths, limitations, and the best ways to improve productivity with them. Building on that interest, I developed services using AI Agents and RAG, and validated those applied skills through hackathon awards.

I have steadily deepened my understanding of finance, including earning a Certified Investment Manager qualification. This background helps me understand financial terminology and business logic in fintech more quickly and accurately. I now put that strength into practice by building and operating a quantitative investment analysis platform covering more than 8,500 stocks across four Korean and U.S. markets.

Spring Boot is my main stack, and I can also use Node.js and Python flexibly depending on the needs of the project. Through about two years of hands-on development experience, including an overseas internship and startup work, I have become comfortable handling React frontend as well when needed. I have also consistently backed up my work with results, including winning 1st place in a major Google-backed hackathon against around 150 participants.

About

Spring Boot Backend Developer

MAIN LANGUAGES

Java

JavaScript

EDUCATION

Kookmin University

[Major]International Business (KIBS)

[Dual Major]Information Technology

2019-03-01 ~ 2026-02-11

Seoul, Republic of Korea

Hello, I'm Kyoungin Nam — a developer who CAFEs: Cloud · AI-Native · Finance · Engineering.

I build backends for financial services. With a business administration background and a Certified Investment Manager qualification, I bring domain understanding to the code I write.

My flagship project is SaramQuant — a quantitative investment analysis platform I built and run solo, covering 8,500+ stocks across four Korean and U.S. markets. It computes 32 core indicators daily using industry-standard methodologies, designed so that even first-time investors can understand what they see at a glance.

I believe good technology can open quality wealth management to far more people. This idea of financial inclusion is what drives my career, and I want to keep growing as a backend developer in fintech. Feel free to look around — I'd love the chance to connect.

Best regards, Kyoungin Nam

AWARDS

GDGoC KR Hackathon: ONE WAVE

1st Place Award

Google Developers Group

At this hackathon themed around the job market crisis, our 4-person team built 'Barrier Free,' a tailored employment platform for people with disabilities. I led the applicant service backend with Spring Boot (Kotlin) and handled planning and presentation, winning 1st place among 37 teams (~150 participants).

2026-02-07

2026 Sogang University AI Winter Camp: Agent Development Mini Hackathon with n8n

2nd Place Award

Sogang University

Developed 'RivalQ' in a 2-person team — a service where entering a competitor's name triggers AI-driven analysis of their products, pricing, and marketing, with RAG-powered strategy recommendations referencing your own company data. Won 2nd place.

2026-01-27

Interdisciplinary & Convergence Major Completion Essay Contest

2nd Place Award

Kookmin University College of Future Convergence

Wrote an essay about my 14-month journey through Vietnam and the United States, winning 2nd Place. Published on the university's official website and major guidebook.

2025-12-01

The 2nd Global Co-Prosperity Joint Academic Conference

3rd Place Award

Kookmin University

Planned and developed 'Kor-Eco,' a waste sorting assistant app for foreign residents in Korea. Led a 4-person team as team leader and full-stack developer, winning 3rd Place.

2025-11-28

CERTIFICATIONS

FINANCE

Certified Investment Manager

Korea Financial Investment Association (KOFIA)

2026-01-29|26-000116

CLOUD

Microsoft Certified: Azure Data Fundamentals (DP-900)

Microsoft

2024-07-21|838DAD2ABAA6CE35

CLOUD

Microsoft Certified: Azure Fundamentals (AZ-900)

Microsoft

2024-07-14|75B21F93AFBE994D

DATABASE

SQL Developer (SQLD)

Korea Data Agency

2024-06-21|SQLD-053011690

LANGUAGES

KoreanNative

EnglishAdvanced

TOEIC905/990

2024-03-31142943

ChineseBeginner

HSK Level 3271/300

2024-02-04H32402005742

Experience

From US startups to global internships

Mar 2026

Feb 2019

Mar 2025

GALLERY

Real-time call analysis and speaker separation pipeline

I built a sales coaching flow that connected calls, real-time speaker separation, speech-to-text, and AI summaries in one pipeline.

TwiMLWebSocketGoogle STTMedia StreamAudio Processing

•
Call masking system: Using TwiML, I built a call masking feature that allowed two-way calls without exposing the agent’s personal number.
•
Real-time speaker separation and STT: I separated caller and callee audio from Twilio Media Stream and routed each stream through its own WebSocket channel. I also converted μ-law 8kHz audio to PCM 16kHz in real time, so both speakers could be transcribed clearly and with low latency in Google STT.
•
Automatic AI insights after each call: After each call ended, I sent the transcript to Claude so the system could automatically generate useful sales insights such as summaries and customer intent.

Email campaign system with separated server roles

I built a marketing system that supported both HTML and MJML, while splitting server responsibilities to keep the system more stable.

MJMLSendgridLoad Balancing

•
HTML and MJML dual editor: I provided both an HTML code editor for advanced users and an MJML-based visual editor for non-technical users. This made it much easier to build emails with drag-and-drop actions like adding buttons, adjusting images, and arranging layouts.
•
Clear server role separation: I split the workload between servers so Node.js handled MJML-to-HTML rendering, while Spring Boot took care of bulk sending and heavier I/O work. This helped the overall system run more reliably.

Document automation and contract collaboration workflow

I built a flexible system that connected document drafting, review, and signing in one smooth contract workflow.

contentEditableTokenizationPDF ManagementSendGrid

•
Flexible e-signature handling: I supported both typed signature generation and handwritten signature uploads, with image preprocessing to extract only the signature itself.
•
Dynamic contract structure: I made signature sections grow or shrink automatically based on how many buyers or sellers were involved. Fields such as addresses and phone numbers were also handled like variables, so users could edit them directly in a contentEditable-based editor.
•
Token-based review system: I built a review feature where reviewers could highlight exact phrases, leave comments, and assign owners. By tokenizing every word and giving it an ID, the system could still track positions even after the text had been edited.
•
From document generation to approval: I automated the flow from PDF generation with Playwright to email delivery through SendGrid. Recipients could go straight into approval and signing with only an email link and security code, without needing to sign up first.

Event-based system emails and template engine

I designed the system so the right emails would be sent automatically for the different events that happen inside the product.

EJSSendGridEvent-Driven Architecture

•
Event-triggered sending flow: I set up automatic email delivery for common events such as password resets, welcome emails, email OTP verification, and account deactivation.
•
EJS templates built for maintainability: I designed and built the EJS template structure myself so each scenario could bind the right data naturally, making future updates and expansion much easier.

Access control and audit logging architecture

I designed the security foundation step by step, covering authentication, authorization, and audit tracking.

JWT (RS256)Google OAuth2FA (TOTP)RBACBridge TokenAudit Logging

•
RS256 JWT and bridge tokens: I built an RS256-based JWT authentication system and added bridge tokens for intermediate steps such as waiting for 2FA or choosing a company. This made the login flow much more precise and stable.
•
Google OAuth and 2FA: I implemented Google OAuth for convenience and 2FA with OTP to strengthen account security.
•
RBAC designed with ITGC in mind: Based on a zero-trust mindset, I clearly separated permissions for admins, general users, and team members. I also designed the wider flows, like team management, invitations, and permission changes, so controls stayed consistent throughout.
•
Audit logging and access tracking: I built an audit logging system that recorded API activity and resource access for users, admins, and system accounts in the database. It was designed to support security incident analysis and help trace permission misuse.

SMS campaign delivery and cost management system

I built an SMS system that covered bulk sending, retry handling, opt-out processing, and cost calculation.

RedisBullMQTwilio API

•
Async bulk sending and retry handling: Using Redis and BullMQ, I handled bulk sending requests asynchronously and added exponential backoff retries so the system could recover more gracefully from temporary external API issues.
•
Automatic opt-out handling and compliance: I automated opt-out handling with keywords such as STOP and START, which helped the product follow TCPA rules in the U.S. more safely.
•
Real-time cost tracking and billing: I calculated message segments automatically based on GSM-7 and Unicode encoding to track sending cost, then reflected platform markup as well so the billing structure could be shown in real time.

Tech Stack

Spring BootNode.jsNext.jsReactPostgreSQL

GALLERY

Sep 2024

GALLERY

VORI-Remade (voice-based workflow automation)

I refactored a collaboration system that turned meeting audio into real-time transcripts and used AI to create Kanban tickets automatically.

WebSocketOpenAI WhisperKafkaAWS

•
AWS architecture design and business validation: With guidance from an AWS mentor, I designed a distributed architecture using Lambda and MSK. Later, I presented the project at the AWS Irvine office in an investor-style setting and helped validate its business potential.
•
Better speed and lower cost through refactoring: To solve the rising cost caused by managed-service dependency, I redesigned the system structure. By moving to a real-time pipeline based on WebSocket and Whisper, I improved workflow speed by 57% and reduced operating cost by 83%.
•
Lower token cost with hybrid processing: I split responsibilities so the LLM handled complex context understanding, while the backend handled simple statistics and word frequency calculations directly. This reduced token usage in a meaningful way.

MakeEat (AI recipe recommendations from receipts)

I built a service that extracted ingredients from receipt images and used them to recommend recipes and nutrition information.

PyTorch (EasyOCR)Tesseract OCRGPT-4Node.js

•
Hybrid OCR pipeline: I separated text detection from text recognition and applied the right model to each step, which improved recognition accuracy by 56.2% compared to using a single model.
•
Lower cost and better analysis accuracy: I reduced cost by 80.7% by sending preprocessed text to GPT-4 instead of raw images. I also split nutrition analysis into stages and cross-checked with an external database, improving accuracy by 55% over similar services.

Pawsome Day (workflow automation for dog daycare centers)

I planned and built a service for local dog daycare centers in the U.S. that helped automate photo sorting and daily reports.

OpenCVNumPyFastAPIWeb ScrapingTableau

•
Automatic dog photo classification: I preprocessed images with OpenCV, extracted facial features, and connected NumPy-based data handling logic so dog photos could be sorted automatically inside the album flow.
•
Field-focused UX research: I collected and analyzed reviews from 80 dog daycare centers in California and also carried out field interviews, so the product requirements reflected the real pain points of operators.

Tech Stack

FastAPINode.jsAWSPostgreSQL

GALLERY

Jun 2024

GALLERY

Global attendance data integration system

I built an environment where commuting and attendance data from multiple countries could be gathered in one place and analyzed in real time.

Azure Data FactorySQL ServerPower BIVisual StudioSharePoint

•
ETL pipeline: Using Azure Data Factory, I built an ETL pipeline that automatically collected and integrated data scattered across APIs and SharePoint.
•
SQL-based data processing and reporting: I organized the data with SQL Server and Visual Studio, then wrote SRS-based analysis reports so leadership could understand the situation more quickly.
•
Power BI dashboard visualization: I built Power BI dashboards that made company-wide attendance status visible in real time and presented decision-making insights in a more intuitive way.