Sunwoo Lee

Worked on connector patterns for heterogeneous customer directory systems and built safeguards to reduce the risk of large-scale HR data mistakes.

• Implemented HR/directory integrations for 117 customers across LDAP, REST, DB, File, MQ, UDP, and Oracle, MSSQL, PostgreSQL, MySQL, and MariaDB
• Implemented data safety guards for bulk HR updates using reduction-rate, absolute-minimum, and resignation thresholds
• Analyzed bottlenecks and clarified processing flows in Spring Batch-based HR synchronization jobs to improve operational troubleshooting
• Refactored async flows from Future-based logic to CompletableFuture, improving exception handling and maintainability

Tech: Java 21, Spring Boot, Spring Batch, MyBatis, Oracle, MSSQL, PostgreSQL, MySQL, MariaDB, LDAP, REST API, IBM MQ, UDP

Even with the same standard, each IdP implements it differently, which can cause integration issues. Implemented and operated SAML SP configuration and session invalidation flows to handle those differences inside a single product.

• Implemented and operated SAML SP integrations for 24 customers, handling per-customer IdP differences across Okta, OneLogin, MS Entra ID, Ping ID, and Authentik
• Organized configuration values, metadata handling, and operational checks for multi-IdP support
• Implemented SAML SLO-based IdP logout and session invalidation flows, handling operational differences across IdPs
• Investigated and resolved login and permission-related integration issues across customer environments

Tech: Spring Boot, SAML 2.0, Okta, OneLogin, MS Entra ID, Ping ID, Authentik, Access Control

Contributed to the team's Kubernetes access-control roadmap initiative, helping build per-user access control and auditing for AWS EKS cluster access.

• Contributed to AWS IAM AssumeRole-based access control that registers/removes each user's IAM User ARN in the Trust Policy, resolving the account-sharing and lack-of-individual-traceability limits of the ServiceAccount token approach
• Applied IAM Roles Anywhere (X.509 certificate-based STS temporary credentials) to integrate AWS API credentials in air-gapped environments without long-lived access keys
• Contributed to IAM Role provisioning backend (/ASA) and K8S Connector Server integration for collecting cloud accounts and permissions
• Implemented per-user tracking and auditing of kubectl commands over AssumeRole sessions via ApiLoggingFilter integration
• Leading a PoC to unify EKS authentication through Keycloak (OIDC IdP), removing the AWS CLI dependency — evaluating EKS OIDC Identity Provider integration and IAM Roles Anywhere (X.509) (in progress)

Tech: Java, Spring Boot, Kubernetes, AWS EKS, IAM AssumeRole, IAM Roles Anywhere, STS

Related open-source implementation (K8s-native generalization): github.com/snwlee/KubeAccessBroker

Built a notification channel layer that connects each customer's different alert infrastructure (internal messengers, MQ, SMS, etc.) through a single interface.

• Integrated 6 notification channels across 85 customers: Slack, KakaoTalk, Email, SMS, MQ, and UDP
• Analyzed deadlocks in notification processing flows and clarified processing order and failure isolation rules to prevent recurrence
• Defined retry policies by failure type and routed failed notifications to a DLQ, enabling controlled recovery and reprocessing
• Handled per-customer notification customization beyond standard channels — channel selection, message format, and delivery conditions

Tech: Spring Boot, REST API, IBM MQ, UDP

Implemented internal automation pipelines and tooling to reduce recurring work and improve operational visibility.

• Built AWS Bedrock usage monitoring with Kinesis Data Firehose → S3 → Glue → Athena plus CloudWatch, CloudTrail, and SNS alerting
• Led the design and development of an enterprise upgrade automation pipeline — full automation from issue intake through AI code-fix generation, build verification, and result posting
• Intake & classification: Redmine incremental sync (30-min schedule) → pgvector semantic search → 15-type auto-classification → Type1/Type2 dual-strategy decision
• AI impact analysis: GitLab repository auto-discovery and custom patch extraction, then Claude-based analysis (risk score 0–100) with past cases and DB schema version notes injected via RAG
• Review & code generation: Claude-based automated code-fix generation (dry-run diff, line-level review) in a Next.js + FastAPI developer review dashboard
• Build & posting: Docker/Maven build automation with status tracking → automated Redmine comment posting (Textile conversion)
• Implemented a RAG-based internal technical support chatbot combining function-level code indexing (tree-sitter + pgvector) with issue, source, and Notion document search, plus automated meeting-briefing generation from tickets (including attachment OCR and citation mapping)
• Built a TypeScript CLI that assembles per-deployment context into customer-specific NotebookLM chatbots, and operated an internal AI newsletter pipeline for the engineering team

Tech: Python, FastAPI, Next.js, PostgreSQL, pgvector, sentence-transformers, tree-sitter, GitLab API, Redmine API, Notion API, Claude Code, AWS Bedrock, Kinesis Data Firehose, S3, Glue, Athena, Docker, Maven, TypeScript, n8n

Automated and documented recurring engineering tasks across security checks, environment upgrades, and test infrastructure.

• Integrated Sparrow into the CI/CD pipeline to reduce manual security scan runs per project
• Helped with SVN → Git migration by organizing branch strategy and commit conventions
• Automated JDK 8 → 21 upgrades and sensitive property migration to KMS in reusable form
• Configured Docker-based Oracle, MSSQL, PostgreSQL, MySQL, and MariaDB test environments

Focus: CI/CD, Git Migration, Java Upgrade Automation, KMS, Docker