Kim Seogyu

Senior Software Engineer

Financial Systems & Data Integrity Specialist

한국어
asap0208@gmail.com +82-10-2400-4572

Professional Summary

"Senior Software Engineer specializing in Distributed Systems and Financial Transaction Reliability. Proven track record of designing high-throughput storage (1,500 TPS) and ensuring 99.999% data integrity. Experienced in bridging Blockchain technology with Enterprise Finance."

Core Engineering
Designed high-performance distributed storage achieving 1,500 TPS per instance with 99.999% data integrity.
LLM & RAG
Built Hybrid RAG pipeline with 95% risk detection reliability (Kiwi + BM25 + FAISS).
Transaction Reliability
Reduced financial transaction failure rate from 10% to <0.1% through idempotency design and ledger reconciliation.

Technical Skills

Languages
GoRustPythonTypeScript (NestJS)SQL
Backend & Infrastructure
Kubernetes (EKS)gRPCKafkaRedis StreamsMongoDBPostgreSQLTerraformHelm
Blockchain & Core
Ethereum (EVM)SolidityMerkle TrieCelestia (DA)Smart Contracts
AI & Data
RAG PipelineLangChainFAISS (VectorDB)LLM Integration

Work Experience

KB Securities

Sep 2025 – Present
Senior AI Solution Architect | AI Tech TeamSeoul, South Korea

KB AI Agent Platform

Spearheaded the development of an Enterprise RAG Core Engine and Agent Workflow for automated risk review of complex financial documents.

PythonLangGraphFastAPIPostgreSQLFAISSpymupdf
  • Advanced Document ParsingEngineered 2-Pass Hierarchy Strategy (Candidate Extraction + Global Context) and Index Snapping algorithm to resolve complex nested structures in financial agreements.
  • Smart Table StitchingImplemented PDF Table Stitching pipeline using pymupdf-layout to merge page-spanning tables and correct deeply nested HTML parsing errors.
  • Robust Agent WorkflowDesigned Adaptive Filtering Logic and Dynamic Batch Query Generation (Batch Size=10) to minimize hallucination and optimize token usage.
  • Self-Correction PipelineBuilt an automated OCR Fallback mechanism to handle text extraction failures (FileEmptyException) by switching to vision-based processing/cleaning.
  • Backend EngineeringDeveloped Secure Stream Processing for efficient large-file handling (bypassing memory limits) and a high-concurrency Async API architecture.
Impact:Achieved significant improvement in parsing accuracy for structured financial data and ensured 99.9% system stability during large-scale batch processing.

42dot (Hyundai Motor Group)

May 2023 – Sep 2025
Backend Engineer | Blockchain Platform TeamSeoul, South Korea

Off-chain Distributed Storage (offchain-db-v3)

Large-scale off-chain storage solution overcoming blockchain cost/performance limitations

GoMongoDBRedis StreamsgRPCKubernetes
  • Designed high-performance storage architectureusing Reed-Solomon Erasure Coding (based on Celestia/Ceph research).
  • Implemented custom in-memory Merkle Patricia Trie(based on Go-Ethereum) to verify batch audit log integrity with single root hash.
  • Built robust event pipelineusing Redis Streams with At-least-once Delivery and Dead Letter Queues.
Impact:Achieved 1,500 TPS per instance and 99.999% data integrity with high availability.

Enterprise Blockchain Indexer SDK

Rust SDK for high-speed blockchain indexing via ETL pipeline

RustgRPCPostgreSQLDieselTokio
  • Architected generic ETL enginein Rust (gRPC streaming extraction, BCS decoding, chunk-based loading).
  • Implemented History/Snapshot dual-table strategysupporting both O(1) real-time state lookups and historical auditing.
  • Abstracted complex chain logicinto Rust Traits, reducing boilerplate code by 90% for internal developers.

Coinone

Feb 2022 – May 2023
Software Engineer | Wallet TeamSeoul, South Korea

Financial Transaction Reliability

Node.jsTypeScript (NestJS)PostgreSQLRedis
  • Designed Idempotency keysand stabilized retry logic to prevent duplicate transactions.
  • Automated Ledger Reconciliation pipelinesto verify data consistency between on-chain and internal databases.
  • Managed node infrastructurefor 10+ blockchains (Bitcoin, Ethereum, etc.) and developed wallet services.
Impact:Reduced deposit/withdrawal failure rate from 10% to <0.1% and cut manual operations by 90%.

Staking Service 'Plus 2.0' Architecture Review

Node.jsTypeScript (NestJS)PostgreSQL
  • Conducted emergency architectural reviewpre-launch, redesigning core modules to resolve critical bottlenecks.
Impact:Achieved zero downtime and zero critical incidents during launch period.

Axiasoft → The Future Company

Mar 2021 – Feb 2022
Backend Developer | Metaverse PlatformSeoul, South Korea

Metaverse Virtual Land Trading System

Led MVP development from architecture to deployment for global virtual real estate platform

Node.js (NestJS)PostgreSQL (PostGIS)Turf.jsRedis
  • Optimized spatial data processingusing PostGIS and Turf.js for R-tree indexing and polygon operations.
  • Implemented secure payment systemsintegrating PayPal/Coocon and OAuth 2.0 based authentication.
Impact:Successfully launched MVP, contributing to spinoff and new corporate entity establishment.

Hongik University (Visual Communication Design)

Aug 2020 – Mar 2021
Backend Developer (Freelance)Seoul, South Korea

Official Department Website (sidi.hongik.ac.kr)

PythonDjangoAWS
  • Developed CMSusing Python (Django) and AWS (EC2, S3) for archiving portfolios and managing exhibitions.

Education

42 Seoul

Software Engineering (Peer-to-Peer Learning Program)

Jan 2020 – Jan 2021

Kyonggi University

B.A. in Korean Language & Literature / Business Administration (Double Major)

Mar 2013 – Feb 2020