cho.sh
2,238 notes
261021261021Toward Autopoiesis26F6DF오토포이에시스를 향해84A5EE차세대 디지털 브레인 만들기C24273261019261019261018261018디지털 시대의 언어 다양성을 위한 선언문507C2CManifesto for Linguistic Diversity in the Digital Age6B5B52261012261012코리지빌리티1CAB16261005261005261004261004Mesh EngineeringF27D1B그물망 엔지니어링881E3BThe devil is in the details94F308260923260923260919260919260917260917In the Agentic Era, should we become a dumb pipe?D815AC260916260916Coscientist6975BA260915260915Coscientific Swarm5F70E8협력적 연구 군체B9E26A
AI-native

AI-native

6-people team

RoleConsolidated ScopePrimary OwnershipExplicit Boundary
1. AI Product & UX EngineerProduct, UX, frontendUser journeys, interaction design, feedback UX, human-in-the-loop controls, frontend implementationDoes not own orchestration, model behavior, or infrastructure
2. AI Application EngineerAI application backendAgent workflows, prompts, tool calling, memory, business logic, application APIsConsumes retrieval and model APIs but does not own them
3. Knowledge & Data EngineerData engineering, search, RAGData ingestion, document processing, embeddings, indexing, retrieval, reranking, data qualityOwns context retrieval, not final answer generation
4. Model EngineerModel development and adaptationModel selection, fine-tuning, LoRA, distillation, synthetic data, training pipelinesOwns model weights and behavior, not production serving
5. Inference & Platform EngineerServing, infrastructure, SREModel serving, GPU infrastructure, deployment, batching, caching, scaling, latency, uptime, operational observabilityOwns runtime performance, not answer quality
6. AI Quality, Safety & Security EngineerEvaluation, governance, AI securityEvaluation suites, regression tests, red teaming, guardrails, prompt-injection testing, privacy controls, release gatesDefines and verifies quality and safety requirements, and other roles implement required changes

3-people team

RoleConsolidated ScopePrimary Ownership
1. AI Product & Application EngineerProduct, UX, frontend, application backend, agents, prompts, tools, memoryEverything from user interaction to the model and retrieval API boundary
2. Intelligence & Quality EngineerData, RAG, model selection, fine-tuning, evaluation, safety policyEverything determining what the system knows, how it behaves, and how output quality is measured
3. Inference, Platform & Reliability EngineerModel serving, infrastructure, deployment, scaling, security enforcement, operational observabilityEverything required to run the system securely, reliably, and economically in production

2023

Especially with the emergence of ChatGPT.

AI-native means building and shipping AI the way software gets built. Cloud infrastructure, services, and tools all the way through, to build, train, and deploy models.

  • Data lives in the cloud. Data lakes, data warehouses, and analytics platforms hold the huge piles of data, and cloud services do the storage, processing, and analysis.
  • Models get built with cloud tools. ML frameworks, NLP services, and computer vision APIs, all hosted.
  • Training eats a lot of data from a lot of sources. That data usually sits in the cloud, and the models reach it through cloud services and tools.
  • Models get deployed to the cloud too, so they can scale up and down as the load changes.

What you get is faster time to market, better scalability and availability, and lower cost. What it asks for is people who know data science, ML, and cloud computing. And it comes with its own problems. Handling huge volumes of data, keeping that data private and secure, and keeping bias out of the models.

Backlinks5