📚 Referências

Bibliografia organizada por tema. Cada entrada inclui, quando aplicável:

• Título.
• Autor / Organização.
• Tipo - paper, spec, docs, blog técnico oficial, framework, benchmark, livro.
• Por que é relevante - em uma linha.
• Seção do livro - onde a referência é usada.

Princípio editorial. Documentação de fornecedor está incluída apenas quando é a melhor fonte técnica disponível. Material puramente promocional não aparece como referência principal; quando útil, fica em "Leitura complementar" claramente identificada. Tópicos emergentes têm o nível de maturidade declarado explicitamente, e o livro não afirma consenso onde ainda não há.

---

🗂️ Sumário

• 📚 Referências
• 🗂️ Sumário
• 📊 1. ML tradicional, MLOps e estatística
  • 📑 Papers e livros
  • 📘 Documentação oficial
  • 📜 Blogs técnicos
• 🤖 2. Fundamentos de LLMs, tokenização e fine-tuning
  • 📑 Papers
  • 📘 Documentação oficial e frameworks
  • 📜 Blogs técnicos oficiais
• 🧬 3. Embeddings, Matryoshka e retrieval
  • 📑 Papers
  • 📘 Documentação oficial
  • 📜 Blogs / GraphRAG
• 🚀 4. Serving, throughput e otimização de inferência
  • 📑 Papers
  • 📘 Documentação oficial
• 🧩 5. Agentic AI, MCP e A2A
  • 📘 MCP - Model Context Protocol
  • 🤝 A2A - Agent2Agent
  • 📘 Frameworks e SDKs
  • 📑 Memória em agentes
  • 🏗️ Padrões arquiteturais
• 🛡️ 6. Segurança, identidade e governança
  • 📘 Frameworks de risco
  • 🧰 Engines de política (policy-as-code)
  • 🔐 Identidade e autorização (OAuth/OIDC)
  • 🔓 Prompt injection e guardrails (pesquisa e prática)
• 🏗️ 7. Supply chain de IA
• 🌐 8. AI gateways
• 🔭 9. Observabilidade
• 🔄 10. Workflows e infraestrutura
• 💰 11. FinOps
• ✅ 12. Avaliação
• 🏋️ 13. Treinamento distribuído e escala (vocabulário)
• 🧰 14. Frameworks e ecossistema
• 📖 15. Livros e leituras transversais
• 🚫 Observação sobre fontes promocionais

---

📊 1. ML tradicional, MLOps e estatística

📑 Papers e livros

• Mitchell et al. - Model Cards for Model Reporting (FAT* 2019). Paper original - https://arxiv.org/abs/1810.03993. Base para Model Cards. Usado em Cap. 1.7 e 1.11.
• Gebru et al. - Datasheets for Datasets (CACM 2021). Paper -https://arxiv.org/abs/1803.09010. Base para Dataset Cards. Cap. 1.2, 1.7.
• Hardt, Price & Srebro - Equality of Opportunity in Supervised Learning (NIPS 2016). Paper - https://arxiv.org/abs/1610.02413. Base para equalized odds e equality of opportunity. Cap. 1.11.
• Guo et al. - On Calibration of Modern Neural Networks (ICML 2017). Paper - https://arxiv.org/abs/1706.04599. Calibração e ECE. Cap. 1.3.
• Ribeiro, Singh & Guestrin - "Why Should I Trust You?" Explaining the Predictions of Any Classifier (KDD 2016). Paper -https://arxiv.org/abs/1602.04938. LIME. Cap. 1.11.
• Friedman - Greedy Function Approximation: A Gradient Boosting Machine (Annals of Statistics 2001). Paper -https://projecteuclid.org/journals/annals-of-statistics/volume-29/issue-5/Greedy-function-approximation-A-gradient-boosting-machine/10.1214/aos/1013203451.full. Origem do Partial Dependence Plot. Cap. 1.11.
• Barocas, Hardt & Narayanan - Fairness and Machine Learning: Limitations and Opportunities (livro aberto). https://fairmlbook.org/. Cap. 1.11.
• Chip Huyen - Designing Machine Learning Systems (O'Reilly). Livro. Cobre pipelines, monitoramento e deployment. Cap. 1.1-1.10.
• Lakshmanan, Robinson & Munn - Machine Learning Design Patterns (O'Reilly). Livro. Padrões reutilizáveis. Cap. 1.5-1.10.

📘 Documentação oficial

• scikit-learn - https://scikit-learn.org/stable/. Docs (framework).
• PyTorch - https://pytorch.org/docs/stable/index.html. Docs (framework).
• JAX - https://docs.jax.dev/. Docs (framework).
• MLflow - https://mlflow.org/docs/latest/index.html. Docs (experiment tracking + registry). Cap. 1.7.
• Kubeflow - https://www.kubeflow.org/docs/. Docs (orchestration).
• Feast - https://docs.feast.dev/. Docs (feature store).
• Feast - Point-in-time joins -https://docs.feast.dev/getting-started/concepts/point-in-time-joins. Cap. 1.5.
• Tecton - https://docs.tecton.ai/. Docs (feature store comercial).
• Evidently AI - https://docs.evidentlyai.com/. Docs (drift, monitoring). Cap. 1.2, 1.9.
• Great Expectations - https://docs.greatexpectations.io/. Docs (data quality). Cap. 1.2.
• Pandera - https://pandera.readthedocs.io/. Docs (validation). Cap. 1.2.
• Dagster - https://docs.dagster.io/. Docs (orchestration).
• Apache Airflow - https://airflow.apache.org/docs/apache-airflow/stable/. Docs (orchestration).
• SHAP - https://shap.readthedocs.io/. Docs (interpretabilidade). Cap. 1.11.

📜 Blogs técnicos

• Evidently AI - What is concept drift in ML -https://www.evidentlyai.com/ml-in-production/concept-drift. Cap. 1.9.
• Google Developers - Classification: accuracy, precision, recall -https://developers.google.com/machine-learning/crash-course/classification/accuracy-precision-recall. Cap. 1.3.
• ICLR Blog 2025 - Understanding model calibration (ECE) -https://iclr-blogposts.github.io/2025/blog/calibration/. Cap. 1.3.
• Sebastian Raschka - LoRA vs full fine-tuning -https://sebastianraschka.com/faq/docs/lora-vs-full-finetuning.html. Cap. 2.2.
• Hugging Face - Tokenizers -https://huggingface.co/docs/tokenizers/en/index. Cap. 2.1.

---

🤖 2. Fundamentos de LLMs, tokenização e fine-tuning

📑 Papers

• Vaswani et al. - Attention Is All You Need (NeurIPS 2017). Paper -https://arxiv.org/abs/1706.03762. Transformer fundacional. Cap. 2.1.
• Sennrich, Haddow & Birch - Neural Machine Translation of Rare Words with Subword Units (ACL 2016). Paper - https://aclanthology.org/P16-1162/. Origem do BPE. Cap. 2.1.
• Kudo & Richardson - SentencePiece: A simple and language independent subword tokenizer (EMNLP 2018 demo). Paper -https://arxiv.org/abs/1808.06226. Cap. 2.1.
• Hu et al. - LoRA: Low-Rank Adaptation of Large Language Models (ICLR 2022). Paper - https://arxiv.org/abs/2106.09685. Cap. 2.2.
• Dettmers et al. - QLoRA: Efficient Finetuning of Quantized LLMs (NeurIPS 2023). Paper - https://arxiv.org/abs/2305.14314. NF4, double quantization, paged optimizers. Cap. 2.2.
• Frantar et al. - GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers (ICLR 2023). Paper -https://arxiv.org/abs/2210.17323. Cap. 2.2.
• Lin et al. - AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration (MLSys 2024). Paper -https://arxiv.org/abs/2306.00978. Cap. 2.2.
• Xiao et al. - SmoothQuant (ICML 2023). Paper -https://arxiv.org/abs/2211.10438. Cap. 2.2.

📘 Documentação oficial e frameworks

• OpenAI API - https://developers.openai.com/api/docs. Docs.
• OpenAI - Function calling -https://developers.openai.com/api/docs/guides/function-calling. Cap. 2.4, 3.5.
• OpenAI - Structured outputs -https://developers.openai.com/api/docs/guides/structured-outputs. Cap. 0.2, 2.4.
• OpenAI - Prompt caching -https://developers.openai.com/api/docs/guides/prompt-caching. Cap. 2.3, 6.4.
• OpenAI - Latency optimization -https://developers.openai.com/api/docs/guides/latency-optimization. Cap. 6.3, 6.4.
• OpenAI - Supervised fine-tuning -https://developers.openai.com/api/docs/guides/supervised-fine-tuning. Cap. 2.2.
• OpenAI - Working with evals -https://developers.openai.com/api/docs/guides/evals. Cap. 2.11, 5.4.
• OpenAI - Evaluation best practices -https://developers.openai.com/api/docs/guides/evaluation-best-practices. Cap. 5.4.
• OpenAI - Evaluate agent workflows -https://developers.openai.com/api/docs/guides/agent-evals. Cap. 5.4.
• OpenAI - Graders -https://developers.openai.com/api/docs/guides/graders. Cap. 5.4.
• Anthropic Claude docs - https://docs.anthropic.com/. Docs.
• Anthropic - Tool use -https://docs.anthropic.com/en/docs/build-with-claude/tool-use. Cap. 2.4, 3.5.
• Anthropic - Prompt caching -https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching. Cap. 6.4.
• Google ADK - https://adk.dev/. Docs.
• Google Vertex AI - https://docs.cloud.google.com/vertex-ai/docs. Docs.
• Microsoft Foundry Agent Service -https://learn.microsoft.com/en-us/azure/foundry/agents/overview. Docs.
• Microsoft Agent Framework -https://learn.microsoft.com/en-us/agent-framework/overview/. Docs.
• Semantic Kernel - https://learn.microsoft.com/en-us/semantic-kernel/. Docs.
• Amazon Bedrock - https://docs.aws.amazon.com/bedrock/latest/userguide/. Docs.
• bitsandbytes - https://huggingface.co/docs/bitsandbytes/main/en/index. Docs (8-bit/4-bit em Hugging Face). Cap. 2.2.
• llama.cpp e GGUF - https://github.com/ggml-org/llama.cpp. Projeto (inferência local + formato de quantização). Cap. 2.2.
• MLC LLM - https://llm.mlc.ai/. Docs (inferência cross-hardware).

📜 Blogs técnicos oficiais

• Anthropic - Effective context engineering for AI agents -https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents. Cap. 2.3, 3.3.
• Anthropic - Demystifying evals for AI agents -https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents. Cap. 5.4.
• Anthropic - Effective harnesses for long-running agents -https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents. Cap. 3.3.
• Anthropic - Writing effective tools for agents -https://www.anthropic.com/engineering/writing-tools-for-agents. Cap. 3.5.
• Anthropic - Code execution with MCP -https://www.anthropic.com/engineering/code-execution-with-mcp. Cap. 4.10.
• Anthropic - Measuring AI agent autonomy in practice -https://www.anthropic.com/news/measuring-agent-autonomy. Cap. 3.2.
• OpenAI Cookbook - Prompt Caching 201 -https://developers.openai.com/cookbook/examples/prompt_caching_201. Cap. 2.3.
• BentoML - Prefill–decode disaggregation -https://bentoml.com/llm/inference-optimization/prefill-decode-disaggregation. Cap. 2.1, 6.3.

---

🧬 3. Embeddings, Matryoshka e retrieval

📑 Papers

• Kusupati et al. - Matryoshka Representation Learning (NeurIPS 2022). Paper - https://arxiv.org/abs/2205.13147. Base teórica de Matryoshka. Cap. 2.8.
• Malkov & Yashunin - Efficient and robust approximate nearest neighbor search using HNSW (TPAMI 2018). Paper - https://arxiv.org/abs/1603.09320. Cap. 2.7, 2.8.
• Lewis et al. - Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (NeurIPS 2020). Paper - https://arxiv.org/abs/2005.11401. Cap. 2.6.
• Edge et al. (Microsoft Research) - From Local to Global: A Graph RAG Approach to Query-Focused Summarization. Paper -https://arxiv.org/abs/2404.16130. Cap. 2.9.

📘 Documentação oficial

• OpenAI - Embeddings (inclui dimensions parameter) -https://platform.openai.com/docs/guides/embeddings. Cap. 2.8.
• Cohere - Embed v3 -https://docs.cohere.com/docs/cohere-embed. Suporte a Matryoshka. Cap. 2.8.
• Nomic - Nomic Embed Text -https://docs.nomic.ai/atlas/embeddings-and-retrieval/text-embedding. Modelo Matryoshka open-weight. Cap. 2.8.
• Sentence Transformers - Matryoshka embeddings -https://www.sbert.net/examples/training/matryoshka/README.html. Cap. 2.8.
• Faiss - https://faiss.ai/. Biblioteca ANN. Cap. 2.7, 2.8.
• Cormack, Clarke & Buettcher - Reciprocal Rank Fusion outperforms Condorcet and individual Rank Learning Methods (SIGIR 2009). Paper -https://plg.uwaterloo.ca/~gvcormac/cormacksigir09-rrf.pdf. RRF em hybrid search. Cap. 2.7.
• Robertson & Zaragoza - revisão BM25 (referência clássica citada no livro). Cap. 2.7.
• Pinecone - Hybrid search -https://docs.pinecone.io/guides/search/hybrid-search. Cap. 2.7.
• Weaviate - Hybrid search -https://docs.weaviate.io/weaviate/search/hybrid. Cap. 2.7.
• Qdrant - Hybrid queries -https://qdrant.tech/documentation/search/hybrid-queries/. Cap. 2.7.
• OpenSearch - Hybrid search -https://docs.opensearch.org/latest/vector-search/ai-search/hybrid-search/index/. Cap. 2.7.
• Elasticsearch - RAG -https://www.elastic.co/docs/solutions/search/rag.
• Milvus - https://milvus.io/docs.
• pgvector - https://github.com/pgvector/pgvector.
• Amazon Bedrock - Reranking -https://docs.aws.amazon.com/bedrock/latest/userguide/rerank.html. Cap. 2.7.
• Cohere - Rerank -https://docs.cohere.com/docs/rerank-overview. Cap. 2.7.

📜 Blogs / GraphRAG

• Microsoft GraphRAG - https://microsoft.github.io/graphrag/. Cap. 2.9.
• Neo4j GraphRAG (Python) -https://neo4j.com/docs/neo4j-graphrag-python/current/. Cap. 2.9.
• Google Cloud - GraphRAG with Spanner Graph -https://docs.cloud.google.com/architecture/gen-ai-graphrag-spanner. Cap. 2.9.
• Toloka - RAG evaluation -https://toloka.ai/blog/rag-evaluation-a-technical-guide-to-measuring-retrieval-augmented-generation/. Cap. 2.6.
• Google Cloud - Retrieval-Augmented Generation -https://cloud.google.com/use-cases/retrieval-augmented-generation. Cap. 2.6.
• Ragas - Metrics -https://docs.ragas.io/en/stable/concepts/metrics/available_metrics/. Cap. 2.11, 5.4.
• TruLens - RAG triad of metrics -https://www.trulens.org/getting_started/core_concepts/rag_triad/. Cap. 2.6, 2.11 (leitura complementar).

---

🚀 4. Serving, throughput e otimização de inferência

📑 Papers

• Kwon et al. - Efficient Memory Management for Large Language Model Serving with PagedAttention (SOSP 2023). Paper -https://arxiv.org/abs/2309.06180. Base do vLLM. Cap. 6.3.
• Leviathan, Kalman & Matias - Fast Inference from Transformers via Speculative Decoding (ICML 2023). Paper -https://arxiv.org/abs/2211.17192. Cap. 6.3.
• Pope et al. - Efficiently Scaling Transformer Inference (MLSys 2023). Paper - https://arxiv.org/abs/2211.05102. Cap. 6.3.

📘 Documentação oficial

• vLLM - https://docs.vllm.ai/en/latest/. Docs (PagedAttention, continuous batching, prefix caching). Cap. 6.3.
• vLLM - Architecture overview -https://docs.vllm.ai/en/latest/design/arch_overview.html. Cap. 6.3.
• NVIDIA TensorRT-LLM - https://nvidia.github.io/TensorRT-LLM/. Docs (in-flight batching, paged KV cache). Cap. 6.3.
• NVIDIA Triton Inference Server -https://docs.nvidia.com/deeplearning/triton-inference-server/user-guide/docs/. Docs. Cap. 1.8.
• Hugging Face TGI -https://huggingface.co/docs/text-generation-inference. Docs. Cap. 6.3.
• Ray Serve - https://docs.ray.io/en/latest/serve/index.html. Docs. Cap. 1.8.
• TensorFlow Serving - https://www.tensorflow.org/tfx/guide/serving. Docs. Cap. 1.8.
• TorchServe - https://pytorch.org/serve/. Docs. Cap. 1.8.

---

🧩 5. Agentic AI, MCP e A2A

📘 MCP - Model Context Protocol

• MCP - Architecture overview -https://modelcontextprotocol.io/docs/learn/architecture. Spec. Cap. 3.5.
• MCP - Authorization (draft) -https://modelcontextprotocol.io/specification/draft/basic/authorization. Spec. Cap. 4.8.
• MCP - Tools spec -https://modelcontextprotocol.io/specification/2025-11-25/server/tools. Spec. Cap. 3.5.
• MCP - Security best practices -https://modelcontextprotocol.io/docs/tutorials/security/security_best_practices. Cap. 3.5, 4.

🤝 A2A - Agent2Agent

• A2A Protocol - site oficial - https://a2a-protocol.org/latest/. Spec emergente. Cap. 3.5.
• A2A Protocol - anúncio do Google (contexto histórico) -https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/. Contexto. Cap. 3.5.
• A2A - Agent Card schema (na spec acima). Cap. 3.5.

Maturidade: tópico emergente (spec evoluindo ao longo de 2024–2026). Tratar como tecnologia em consolidação; sem consenso fechado.

📘 Frameworks e SDKs

• LangGraph - https://docs.langchain.com/oss/python/langgraph/overview.
• LangGraph - Persistence -https://docs.langchain.com/oss/python/langgraph/persistence. Cap. 3.6.
• LangChain - Context engineering -https://docs.langchain.com/oss/python/langchain/context-engineering.
• LangChain - Human-in-the-loop -https://docs.langchain.com/oss/python/langchain/human-in-the-loop.
• LangSmith - Observability -https://www.langchain.com/langsmith/observability.
• LlamaIndex - https://developers.llamaindex.ai/python/framework/.
• AutoGen - https://microsoft.github.io/autogen/dev/.
• OpenAI Agents SDK - Tracing -https://openai.github.io/openai-agents-python/tracing/.
• OpenAI Agents SDK - Guardrails -https://openai.github.io/openai-agents-python/guardrails/.
• OpenAI Agents SDK - Human-in-the-loop -https://openai.github.io/openai-agents-python/human_in_the_loop/.
• Google Vertex AI Agent Engine - Memory Bank -https://docs.cloud.google.com/agent-builder/agent-engine/use/adk.
• Amazon Bedrock Agents -https://docs.aws.amazon.com/bedrock/latest/userguide/agents.html.
• Amazon Bedrock Knowledge Bases -https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html.

📑 Memória em agentes

• Packer et al. - MemGPT: Towards LLMs as Operating Systems. Paper -https://arxiv.org/abs/2310.08560. Cap. 3.6.
• Yao et al. - ReAct: Synergizing Reasoning and Acting in Language Models. Paper - https://arxiv.org/abs/2210.03629.
• Schick et al. - Toolformer: Language Models Can Teach Themselves to Use Tools. Paper - https://arxiv.org/abs/2302.04761.

🏗️ Padrões arquiteturais

• Google Cloud - Choose a design pattern for your agentic AI system -https://docs.cloud.google.com/architecture/choose-design-pattern-agentic-ai-system. Cap. 3.4.

---

🛡️ 6. Segurança, identidade e governança

📘 Frameworks de risco

• OWASP - Top 10 for LLM Applications -https://owasp.org/www-project-top-10-for-large-language-model-applications/. Framework. Cap. 4.
• OWASP - Top 10 for LLM Applications 2025 PDF -https://owasp.org/www-project-top-10-for-large-language-model-applications/assets/PDF/OWASP-Top-10-for-LLMs-v2025.pdf.
• NIST - AI Risk Management Framework 1.0 -https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf. Cap. 0.3, 4.
• NIST - Generative AI Profile (AI RMF) -https://www.nist.gov/itl/ai-risk-management-framework. Cap. 4.2.
• NIST - publicação alternativa do GenAI Profile -https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence.
• MITRE ATLAS - https://atlas.mitre.org/. Threat catalog. Cap. 4.2.
• Microsoft - STRIDE Threat Model (ferramenta e vocabulário) -https://learn.microsoft.com/en-us/azure/security/develop/threat-modeling-tool-threats. Cap. 4.2.

🧰 Engines de política (policy-as-code)

• Open Policy Agent - Policy Language (Rego) -https://www.openpolicyagent.org/docs/policy-language. Cap. 4.7.
• Open Policy Agent - Policy Testing -https://www.openpolicyagent.org/docs/policy-testing. Cap. 4.7.
• AWS Cedar - Policy Language Reference -https://docs.cedarpolicy.com/policies/syntax-policy.html. Cap. 4.7.
• AuthZed SpiceDB - Concepts -https://authzed.com/docs/spicedb/concepts/zedtokens. Cap. 4.7.
• Google Zanzibar (paper) -https://research.google/pubs/pub48190/. Base teórica de SpiceDB.

🔐 Identidade e autorização (OAuth/OIDC)

• IETF RFC 6749 - OAuth 2.0 Authorization Framework -https://datatracker.ietf.org/doc/html/rfc6749. Cap. 4.8.
• IETF RFC 8693 - OAuth 2.0 Token Exchange -https://datatracker.ietf.org/doc/html/rfc8693. Cap. 4.8.
• IETF RFC 8628 - OAuth 2.0 Device Authorization Grant -https://datatracker.ietf.org/doc/html/rfc8628. Cap. 4.8.
• IETF RFC 7636 - PKCE -https://datatracker.ietf.org/doc/html/rfc7636. Cap. 4.8.
• IETF RFC 9700 - OAuth 2.0 Security Best Current Practice -https://datatracker.ietf.org/doc/html/rfc9700. Cap. 4.8.
• OpenID Connect Core 1.0 -https://openid.net/specs/openid-connect-core-1_0.html. Cap. 4.8.
• Kubernetes - ServiceAccount Tokens -https://kubernetes.io/docs/tasks/configure-pod-container/configure-service-account/.
• AWS - IAM Roles for Service Accounts -https://docs.aws.amazon.com/eks/latest/userguide/iam-roles-for-service-accounts.html.
• GCP - Workload Identity Federation -https://cloud.google.com/iam/docs/workload-identity-federation.
• Azure - Managed Identities -https://learn.microsoft.com/en-us/azure/active-directory/managed-identities-azure-resources/overview.

🔓 Prompt injection e guardrails (pesquisa e prática)

• Greshake et al. - Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection (2023). Paper -https://arxiv.org/abs/2302.12173. Cap. 4.3.
• Anthropic - Reduce prompt injection -https://docs.anthropic.com/en/docs/test-and-evaluate/strengthen-guardrails/reduce-prompt-injection. Cap. 4.3.
• Simon Willison - série Prompt injection -https://simonwillison.net/series/prompt-injection/. Cap. 4.3 (vocabulário e casos).

---

🏗️ 7. Supply chain de IA

• SLSA - Supply-chain Levels for Software Artifacts, spec v1.2 -https://slsa.dev/spec/v1.2/. Cap. 4.9.
• SLSA - visão geral - https://slsa.dev/.
• SPDX (Linux Foundation) - https://spdx.dev/. Cap. 4.9.
• SPDX - Especificação - https://spdx.github.io/spdx-spec/.
• CycloneDX (OWASP) - https://cyclonedx.org/. Cap. 4.9.
• CycloneDX - ML-BOM - https://cyclonedx.org/capabilities/mlbom/. Modelos, datasets e hyperparameters. Cap. 4.9.
• OpenSSF - https://openssf.org/. Cap. 4.9.
• OpenSSF Scorecard - https://github.com/ossf/scorecard. Cap. 4.9.
• Sigstore - cosign - https://docs.sigstore.dev/cosign/signing/overview/.
• in-toto - https://in-toto.io/.
• NTIA - SBOM minimum elements - https://www.ntia.gov/SBOM.
• CISA - SBOM resources - https://www.cisa.gov/sbom.
• Hugging Face - Model Cards -https://huggingface.co/docs/hub/model-cards. Cap. 4.9, 1.7.

---

🌐 8. AI gateways

• LiteLLM - LiteLLM AI Gateway (LLM Proxy) -https://docs.litellm.ai/docs/simple_proxy. Open-source. Cap. 2.5.
• Kong - AI Gateway / AI Proxy plugin -https://docs.konghq.com/hub/kong-inc/ai-proxy/. Plugins sobre Kong. Cap. 2.5.
• Portkey - AI Gateway - https://portkey.ai/docs/product/ai-gateway. SaaS + opção self-hosted. Cap. 2.5.
• Cloudflare - AI Gateway -https://developers.cloudflare.com/ai-gateway/. Cap. 2.5.

Postura editorial: comparação neutra. Avaliar licença, providers suportados, modelo de deploy e fit ao seu stack antes de adotar.

---

🔭 9. Observabilidade

• OpenTelemetry - GenAI semantic conventions -https://opentelemetry.io/docs/specs/semconv/gen-ai/. Cap. 2.12, 5.2, 5.3.
• OpenTelemetry - GenAI spans -https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-spans/.
• OpenTelemetry - GenAI agent spans -https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-agent-spans/.
• Langfuse - https://langfuse.com/docs. Observability open-source para LLM/agents (leitura complementar).
• Phoenix / Arize - https://docs.arize.com/phoenix. Open-source (leitura complementar).
• Helicone - https://docs.helicone.ai/. Observability para LLMs (leitura complementar).
• LangSmith - https://www.langchain.com/langsmith/observability. SaaS.
• Grafana Tempo - https://grafana.com/docs/tempo/latest/. Backend de traces (com OTel). Cap. 5.2.
• Jaeger - https://www.jaegertracing.io/docs/. Cap. 5.2.

---

🔄 10. Workflows e infraestrutura

• Temporal - https://docs.temporal.io/.
• Temporal - Workflow Execution -https://docs.temporal.io/workflow-execution.
• AWS Step Functions -https://docs.aws.amazon.com/step-functions/latest/dg/welcome.html.
• AWS Step Functions - Error handling -https://docs.aws.amazon.com/step-functions/latest/dg/concepts-error-handling.html.
• Azure Durable Functions -https://learn.microsoft.com/en-us/azure/azure-functions/durable-functions/durable-functions-overview.
• Apache Airflow - DAGs -https://airflow.apache.org/docs/apache-airflow/stable/core-concepts/dags.html.
• Dagster - https://docs.dagster.io/.

---

💰 11. FinOps

• FinOps Foundation - FinOps Framework -https://www.finops.org/framework/. Cap. 6.5.
• FinOps Foundation - FinOps for AI -https://www.finops.org/wg/finops-for-ai/. Working group, material recente. Cap. 6.5.
• OpenAI - Latency optimization -https://developers.openai.com/api/docs/guides/latency-optimization.
• Anthropic - Prompt caching -https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching.

---

✅ 12. Avaliação

• Anthropic - Demystifying evals -https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents.
• OpenAI - Working with evals -https://developers.openai.com/api/docs/guides/evals.
• OpenAI - Evaluation best practices -https://developers.openai.com/api/docs/guides/evaluation-best-practices.
• OpenAI - Evaluate agent workflows -https://developers.openai.com/api/docs/guides/agent-evals.
• OpenAI - Graders -https://developers.openai.com/api/docs/guides/graders.
• Ragas - Metrics -https://docs.ragas.io/en/stable/concepts/metrics/available_metrics/.
• Ragas - Faithfulness -https://docs.ragas.io/en/stable/concepts/metrics/available_metrics/faithfulness/.
• Ragas - Context Precision -https://docs.ragas.io/en/stable/concepts/metrics/available_metrics/context_precision/.
• Zheng et al. - Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena (2023). Paper - https://arxiv.org/abs/2306.05685. Cap. 2.11, 5.4.
• Liu et al. - G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment (2023). Paper - https://arxiv.org/abs/2211.09110. Cap. 2.11.

---

🏋️ 13. Treinamento distribuído e escala (vocabulário)

Escopo mínimo do Cap. 7.7 — não substitui manuais de cluster GPU.

• Shoeybi et al. - Megatron-LM (tensor parallelism). Paper -https://arxiv.org/abs/1909.08053. Cap. 7.7.
• Rajbhandari et al. - ZeRO: Memory Optimizations Toward Training Trillion Parameter Models. Paper - https://arxiv.org/abs/1910.02054. Cap. 7.7.
• PyTorch - FSDP overview - https://pytorch.org/docs/stable/fsdp.html. Cap. 7.7.
• Microsoft - DeepSpeed - https://www.deepspeed.ai/. Cap. 7.7.
• Hugging Face - Accelerate -https://huggingface.co/docs/accelerate/index. Cap. 7.7.

---

🧰 14. Frameworks e ecossistema

• LangChain - https://docs.langchain.com/.
• LangGraph - https://docs.langchain.com/oss/python/langgraph/overview.
• LangSmith - https://www.langchain.com/langsmith/observability.
• LlamaIndex - https://developers.llamaindex.ai/python/framework/.
• AutoGen - https://microsoft.github.io/autogen/dev/.
• Semantic Kernel - https://learn.microsoft.com/en-us/semantic-kernel/.
• Microsoft Agent Framework -https://learn.microsoft.com/en-us/agent-framework/overview/.
• Google ADK - https://adk.dev/.
• Pinecone - https://docs.pinecone.io/.
• Weaviate - https://docs.weaviate.io/weaviate.
• Milvus - https://milvus.io/docs.
• Qdrant - https://qdrant.tech/documentation/.
• Elasticsearch - https://www.elastic.co/docs/solutions/search/rag.
• OpenSearch - https://docs.opensearch.org/latest/vector-search/.
• pgvector - https://github.com/pgvector/pgvector.

---

📖 15. Livros e leituras transversais

• Chip Huyen - Designing Machine Learning Systems (O'Reilly). Livro.
• Chip Huyen - AI Engineering: Building Applications with Foundation Models (O'Reilly, 2024). Livro.
• Lakshmanan, Robinson & Munn - Machine Learning Design Patterns (O'Reilly). Livro.
• Engineering blogs de OpenAI, Anthropic, Google, Meta, Netflix, Airbnb e Uber. Material variado sobre MLOps, LLMOps e Agentic AI.

---

🚫 Observação sobre fontes promocionais

Quando uma referência é primariamente promocional de um produto, não deve ser usada como única base para uma afirmação técnica. O critério:

• ✅ Documentação oficial com detalhes técnicos verificáveis.
• ✅ Papers revisados ou pré-prints sérios.
• ✅ Especificações abertas (RFCs, MCP, A2A, OpenTelemetry, SLSA, SPDX, CycloneDX).
• ⚠️ Blog post de vendor com substância técnica - citar com cautela e como complementar.
• ❌ Post de marketing puro - não citar.

Tópicos emergentes (A2A, MCP Authorization mais recente, FinOps for AI) têm sua maturidade declarada explicitamente e devem ser revisitados periodicamente.