기술 스택
KGA가 채택한 기술과 그 선정 이유
인프라 & 클라우드
모든 워크로드를 위한 확장 가능한 기반
AWS
Primary cloud provider. We run ECS and Lambda for compute, S3 for object storage, and RDS for managed databases.
Google Cloud Platform
BigQuery for analytics data warehousing and Cloud Run for ML inference workloads that benefit from auto-scaling.
Cloudflare
Global CDN, DDoS protection, and Workers at the edge for low-latency routing and A/B testing logic.
Docker
Every service ships as a container. Consistent environments from local development through production.
Kubernetes
Production orchestration for large-scale client deployments requiring fine-grained scaling and service mesh.
Terraform
Infrastructure as Code for all environments. Reproducible, version-controlled, and peer-reviewed infra changes.
백엔드 & API
안정성을 위해 구축된 고성능 서비스
Node.js / TypeScript
Primary backend runtime for API services, webhooks, and real-time applications. End-to-end type safety with shared types.
Python
AI/ML pipelines, data processing scripts, and automation tooling. Used with FastAPI for model-serving endpoints.
Go
High-performance microservices where latency and throughput matter -- ingestion pipelines, proxy layers, and CLI tools.
FastAPI
ML model serving with automatic OpenAPI docs. Handles inference requests with async support and Pydantic validation.
GraphQL
Flexible API layer for client dashboards. Reduces over-fetching and lets frontend teams iterate without backend changes.
gRPC
Internal service-to-service communication with protobuf schemas. Strong contracts, low overhead, bidirectional streaming.
프론트엔드
사용자가 사랑하는 모던 인터페이스
Next.js
Primary framework for all web applications. Server components, edge rendering, and built-in optimizations.
React / React Native
Shared component libraries across web and mobile. Native iOS and Android apps from a single codebase.
Svelte
Lightweight micro-frontends and embeddable widgets where bundle size is critical.
Tailwind CSS
Foundation of our design system. Utility-first approach ensures consistency and rapid iteration across projects.
Three.js
3D and VR experiences for product configurators, architectural walkthroughs, and interactive data visualizations.
AI & 머신러닝
학습하고 적응하는 지능형 시스템
OpenAI API
GPT-4o for natural language tasks, embeddings for semantic search, and fine-tuned models for domain-specific use cases.
Hugging Face
Open-source model hosting and inference. We run specialized NLP and vision models for clients with data residency requirements.
TensorFlow / PyTorch
Custom model training for computer vision, time-series forecasting, and recommendation engines.
LangChain
AI agent orchestration for complex multi-step workflows -- document Q&A, data extraction, and autonomous task execution.
Pinecone
Vector database powering RAG pipelines. Sub-50ms similarity search across millions of document embeddings.
Ollama
Local LLM deployment for air-gapped environments and clients who need on-premise AI with zero data egress.
데이터 & 데이터베이스
데이터 형태에 맞는 올바른 저장소
PostgreSQL
Primary relational database. ACID transactions, JSONB for flexible columns, and extensions like pgvector for embeddings.
Redis
In-memory caching, session management, rate limiting, and BullMQ job queues for background processing.
MongoDB
Document storage for content management, event sourcing, and schemas that evolve rapidly during product discovery.
InfluxDB
Time-series database for IoT sensor data, application metrics, and real-time dashboards with sub-second resolution.
Elasticsearch
Full-text search across product catalogs and documentation. Also powers centralized log aggregation and analysis.
BigQuery
Serverless data warehouse for analytics. Petabyte-scale SQL queries for business intelligence and reporting.
DevOps & 모니터링
빠르게 배포하고, 모든 것을 관찰
GitHub Actions
CI/CD pipelines for build, test, and deployment. Matrix builds, reusable workflows, and environment-based approvals.
Grafana
Unified observability dashboards combining metrics, logs, and traces. Custom dashboards for every client project.
Prometheus
Metrics collection and alerting. PromQL-powered alerts feed into PagerDuty for on-call incident response.
Sentry
Error tracking with source maps and release tracking. Every unhandled exception gets triaged within minutes.
Datadog
APM and infrastructure monitoring. Distributed tracing across microservices to pinpoint latency bottlenecks.
PagerDuty
Incident management with escalation policies. On-call rotations ensure 24/7 coverage for critical systems.
보안
모든 계층에서의 심층 방어
Vault
Centralized secrets management with dynamic credentials. Database passwords, API keys, and certificates rotate automatically.
Snyk
Continuous dependency scanning in CI. Vulnerabilities are flagged before they reach production.
OWASP ZAP
Automated security testing in staging environments. DAST scans run on every release candidate.
Let's Encrypt
Automated SSL/TLS certificate provisioning and renewal. Every endpoint is encrypted with zero manual intervention.
끊임없이 진화
최고의 솔루션을 제공하기 위해 새로운 기술을 지속적으로 평가하고 채택합니다. 우리의 스택은 의견이 있지만 독단적이지 않으며, 각 문제에 맞는 올바른 도구를 선택합니다.