- Published on
NVIDIA NeMo Guardrails를 사용해 LLM 기반 애플리케이션에 입출력 모더레이션, 토픽 제어, 할루시네이션 감지 등 프로그래밍 가능한 안전장치를 구축하는 방법을 실습합니다.
Llm-safety
All Posts
- 2026-03-03 (57)
- 2026-03 (47)
- deep-learning (27)
- kubernetes (27)
- llm (20)
- devops (20)
- ai-papers (19)
- hadoop (19)
- hbase (15)
- linux (13)
- ai (12)
- gpu (12)
- japanese (12)
- march-2026 (10)
- mlops (10)
- architecture (10)
- database (10)
- english (9)
- transformer (8)
- java (8)
- cuda (8)
- observability (8)
- security (7)
- career (7)
- nvidia (6)
- ai-platform (6)
- agent (6)
- performance (6)
- finance (6)
- optimization (5)
- openai (5)
- langchain (5)
- computer-vision (5)
- pytorch (5)
- kerberos (5)
- python (5)
- docker (5)
- sql (5)
- chatbot (5)
- automation (5)
- postgresql (5)
- monitoring (5)
- cloud-native (5)
- job-interview (5)
- business-japanese (5)
- hospital (5)
- vllm (5)
- image-generation (4)
- rag (4)
- retrospect (4)
- terminal (4)
- opentelemetry (4)
- investment (4)
- spring (4)
- gateway-api (4)
- productivity (4)
- cli (4)
- inference (4)
- generative-ai (4)
- arduino (3)
- iot (3)
- embedded (3)
- robotics (3)
- state-space-model (3)
- linear-attention (3)
- fine-tuning (3)
- mlflow (3)
- model-serving (3)
- langgraph (3)
- kafka (3)
- microservices (3)
- authorization (3)
- hive (3)
- command (3)
- zookeeper (3)
- yarn (3)
- redis (3)
- argocd (3)
- gitops (3)
- container (3)
- logging (3)
- distributed-training (3)
- etf (3)
- keigo (3)
- jlpt (3)
- evening-japanese (3)
- networking (3)
- cka (3)
- serving (3)
- diffusion (3)
- ai-coding (3)
- open-source (3)
- raspberry-pi (2)
- drone (2)
- pid-control (2)
- flight-controller (2)
- gpt (2)
- mamba (2)
- ssm (2)
- moe (2)
- deepseek (2)
- guardrails (2)
- rwkv (2)
- sequence-modeling (2)
- nlp (2)
- pre-training (2)
- google (2)
- generative-model (2)
- stable-diffusion (2)
- rlhf (2)
- lora (2)
- peft (2)
- cnn (2)
- image-classification (2)
- machine-learning (2)
- kubeflow (2)
- experiment-tracking (2)
- model-registry (2)
- workflow (2)
- orchestration (2)
- ml-pipeline (2)
- event-driven (2)
- cqrs (2)
- event-sourcing (2)
- aws (2)
- certification (2)
- cloud (2)
- clean-architecture (2)
- design-pattern (2)
- encryption (2)
- authentication (2)
- oidc (2)
- high-availability (2)
- load-balancing (2)
- tls (2)
- ansible (2)
- grpc (2)
- mysql (2)
- slack (2)
- vector-search (2)
- partitioning (2)
- caching (2)
- ci-cd (2)
- argo-rollouts (2)
- canary (2)
- blue-green (2)
- terraform (2)
- iac (2)
- ros2 (2)
- tensorrt (2)
- sensor-fusion (2)
- hpc (2)
- parallel-computing (2)
- podman (2)
- prometheus (2)
- infrastructure (2)
- distributed-tracing (2)
- evtol (2)
- code-review (2)
- communication (2)
- agile (2)
- business-english (2)
- developer (2)
- it-company (2)
- interview (2)
- tech-career (2)
- evening-english (2)
- business (2)
- irp (2)
- isa (2)
- hdfs (2)
- grammar (2)
- output-training (2)
- migration (2)
- service-mesh (2)
- autoscaling (2)
- cncf (2)
- gpu-operator (2)
- cilium (2)
- ebpf (2)
- kubestronaut (2)
- cks (2)
- kcsa (2)
- tmux (2)
- vim (2)
- neovim (2)
- tool-use (2)
- anthropic (2)
- text-to-video (2)
- image-to-video (2)
- video-generation (2)
- production (2)
- grafana (2)
- loki (2)
- 3d-printing (1)
- fusion-360 (1)
- cad (1)
- modeling (1)
- slicer (1)
- maker (1)
- fdm (1)
- electronics (1)
- nanogpt (1)
- from-scratch (1)
- humanoid-robot (1)
- tesla-optimus (1)
- boston-dynamics (1)
- figure-ai (1)
- unitree (1)
- mathematics (1)
- linear-algebra (1)
- calculus (1)
- probability (1)
- statistics (1)
- gradient-descent (1)
- xiao (1)
- nrf52840 (1)
- ble (1)
- bluetooth (1)
- seeed-studio (1)
- prompt-engineering (1)
- multilingual (1)
- debugging (1)
- mixtral (1)
- nemo-guardrails (1)
llm-safety (1)
- rnn (1)
- attention (1)
- bert (1)
- masked-language-model (1)
- ddpm (1)
- diffusion-model (1)
- score-matching (1)
- latent-diffusion (1)
- flash-attention (1)
- gan (1)
- adversarial-training (1)
- dcgan (1)
- stylegan (1)
- wasserstein (1)
- language-model (1)
- instruction-tuning (1)
- scaling-law (1)
- paper-review (1)
- selective-scan (1)
- mamba2 (1)
- vector-database (1)
- reinforcement-learning (1)
- dqn (1)
- ppo (1)
- a3c (1)
- q-learning (1)
- policy-gradient (1)
- alphago (1)
- resnet (1)
- residual-learning (1)
- skip-connection (1)
- batch-normalization (1)
- sam (1)
- segment-anything (1)
- image-segmentation (1)
- video-segmentation (1)
- meta-ai (1)
- promptable-segmentation (1)
- zero-shot (1)
- foundation-model (1)
- sam2 (1)
- sam3 (1)
- vit (1)
- vision-transformer (1)
- patch-embedding (1)
- self-attention (1)
- feast (1)
- feature-store (1)
- data-engineering (1)
- kfp (1)
- pipeline (1)
- ray-serve (1)
- torchaudio (1)
- audio (1)
- speech-recognition (1)
- spectrogram (1)
- mel (1)
- tts (1)
- music (1)
- torchvision (1)
- object-detection (1)
- segmentation (1)
- transfer-learning (1)
- multi-agent (1)
- state-graph (1)
- tool-calling (1)
- airflow (1)
- golden-jacket (1)
- solutions-architect (1)
- ddd (1)
- domain-driven-design (1)
- spring-boot (1)
- hexagonal-architecture (1)
- ports-adapters (1)
- ipsec (1)
- vpn (1)
- network-security (1)
- ike (1)
- tunnel (1)
- oauth (1)
- jwt (1)
- redundancy (1)
- failover (1)
- disaster-recovery (1)
- sre (1)
- zero-trust (1)
- owasp (1)
- hashing (1)
- xss (1)
- sql-injection (1)
- saga (1)
- distributed-system (1)
- msa (1)
- mongodb (1)
- kdc (1)
- pheonix (1)
- phoenix (1)
- profiler (1)
- crontab (1)
- mapreduce (1)
- api (1)
- scala (1)
- javascript (1)
- raspberry (1)
- spark (1)
- ubuntu (1)
- jvm (1)
- ycsb (1)
- chromadb (1)
- discord-bot (1)
- pycord (1)
- slash-commands (1)
- bolt-sdk (1)
- slack-bot (1)
- telegram (1)
- bot (1)
- webhook (1)
- python-telegram-bot (1)
- fastapi (1)
- deployment (1)
- indexing (1)
- explain (1)
- n-plus-1 (1)
- orm (1)
- django (1)
- sqlalchemy (1)
- jpa (1)
- query-optimization (1)
- cache-aside (1)
- write-through (1)
- redis-cluster (1)
- sharding (1)
- select (1)
- join (1)
- index (1)
- cheatsheet (1)
- pgvector (1)
- embedding (1)
- similarity-search (1)
- hnsw (1)
- ivfflat (1)
- hybrid-search (1)
- github-actions (1)
- self-hosted-runner (1)
- dagger (1)
- pipeline-as-code (1)
- helm (1)
- cicd (1)
- state-management (1)
- applicationset (1)
- progressive-rollout (1)
- multi-cluster (1)
- autonomous-driving (1)
- self-driving (1)
- vla (1)
- vlm (1)
- carla (1)
- simulation (1)
- cpp (1)
- oci (1)
- dockerfile (1)
- containerfile (1)
- buildah (1)
- skopeo (1)
- fluent-bit (1)
- fluentd (1)
- elasticsearch (1)
- keycloak (1)
- iam (1)
- sso (1)
- oauth2 (1)
- nginx (1)
- reverse-proxy (1)
- ssl (1)
- web-server (1)
- rate-limiting (1)
- compression (1)
- otlp (1)
- sampling (1)
- pixhawk (1)
- px4 (1)
- ardupilot (1)
- kalman-filter (1)
- ekf (1)
- pwm (1)
- dshot (1)
- uav (1)
- mavlink (1)
- slurm (1)
- cluster (1)
- job-scheduler (1)
- deepspeed (1)
- ai-infrastructure (1)
- infrastructure-as-code (1)
- configuration-management (1)
- provisioning (1)
- presentation (1)
- tech-english (1)
- public-speaking (1)
- technical-english (1)
- standup-meeting (1)
- email (1)
- meeting (1)
- incident-response (1)
- podcast (1)
- resume (1)
- cover-letter (1)
- it-english (1)
- ats (1)
- linkedin (1)
- hiring (1)
- salary-negotiation (1)
- negotiation (1)
- lunch-english (1)
- tax-saving (1)
- pension (1)
- asset-allocation (1)
- quantitative (1)
- portfolio (1)
- rebalancing (1)
- economics (1)
- financial-literacy (1)
- insurance (1)
- life-insurance (1)
- property-insurance (1)
- actuarial (1)
- risk-management (1)
- savings (1)
- deposit (1)
- stock (1)
- bond (1)
- mixed-precision (1)
- business-email (1)
- it-japanese (1)
- it-career (1)
- tech-writing (1)
- specification (1)
- documentation (1)
- 일본어회화 (1)
- 병원일본어 (1)
- 일본취업 (1)
- it-interview (1)
- engineer (1)
- honorific (1)
- self-introduction (1)
- n5 (1)
- n4 (1)
- n3 (1)
- vocabulary (1)
- kanji (1)
- hiragana (1)
- katakana (1)
- joby-aviation (1)
- 항공택시 (1)
- 주식 (1)
- 투자 (1)
- 사우디-투자 (1)
- 기술 (1)
- ingress (1)
- gamma (1)
- karpenter (1)
- eks (1)
- pod-security (1)
- psa (1)
- pss (1)
- vpa (1)
- in-place-resize (1)
- volcano (1)
- cnpe (1)
- platform-engineering (1)
- crossplane (1)
- backstage (1)
- etcd (1)
- backup (1)
- restore (1)
- hands-on (1)
- kubevirt (1)
- virtualization (1)
- vgpu (1)
- device-plugin (1)
- mig (1)
- ingress-nginx (1)
- http3 (1)
- quic (1)
- nftables (1)
- ipvs (1)
- istio (1)
- ambient-mesh (1)
- cni (1)
- kube-proxy (1)
- rbac (1)
- service-account (1)
- cluster-role (1)
- ckad (1)
- kcna (1)
- rootless (1)
- daemonless (1)
- pasta-networking (1)
- lcel (1)
- nvidia-driver (1)
- multiplexer (1)
- screen (1)
- editor (1)
- ide (1)
- dotfiles (1)
- developer-experience (1)
- qlora (1)
- huggingface (1)
- function-calling (1)
- structured-output (1)
- json-mode (1)
- paged-attention (1)
- ace-step (1)
- music-generation (1)
- ai-music (1)
- audio-generation (1)
- text-to-music (1)
- antigravity (1)
- agentic-ide (1)
- gemini-3 (1)
- claude-code (1)
- developer-tools (1)
- vscode (1)
- cursor (1)
- mla (1)
- deepseek-v3 (1)
- deepseek-r1 (1)
- reasoning (1)
- hunyuan-video (1)
- ltx-video (1)
- ltx2 (1)
- diffusion-transformer (1)
- dit (1)
- tencent (1)
- lightricks (1)
- react (1)
- tensorrt-llm (1)
- codex (1)
- code-generation (1)
- github-copilot (1)
- openclaw (1)
- ai-agent (1)
- mcp (1)
- self-hosted (1)
- sglang (1)
- llm-inference (1)
- radix-attention (1)
- structured-generation (1)
- disaggregation (1)
- speculative-decoding (1)
- draft-model (1)
- token-verification (1)
- latency (1)
- throughput (1)
- text-to-image (1)
- training (1)
- flow-matching (1)
- ollama (1)
- llm-serving (1)
- quantization (1)
- openai-api (1)
- wan (1)
- z-image-turbo (1)
- alibaba (1)
- kserve (1)
- triton (1)
- promtail (1)
- hubble (1)
- tetragon (1)
- otel-collector (1)
- telemetry (1)
- jaeger (1)
- otel (1)
- tempo (1)
- mimir (1)
- tracing (1)
- hbm (1)
- semiconductor (1)
- memory (1)
- ai-hardware (1)
- sk-hynix (1)
- samsung (1)
- micron (1)
- dram (1)
- hbm4 (1)
- data-center (1)