Durable Agents, Disposable Streams
How I separate durable agent execution from disposable live previews with Claude Managed Agents, AppSync Events, and Lambda Web Functions.
Rustem Feyzkhanov · Writing
Notes from building AI platforms and putting agents to work.
Subscribe via RSSHow I separate durable agent execution from disposable live previews with Claude Managed Agents, AppSync Events, and Lambda Web Functions.
How I put OpenAPPA policy checks in front of Claude Managed Agents tool calls, outside the sandbox, with human approval from Web or Slack.
An interaction bridge brings durable coding-agent workflows to Web and Slack while keeping conversation history and work together.
Twenty AWS re:Invent 2024 breakout sessions covering generative AI, models, infrastructure, and production applications.
A step-by-step Mistral 7B deployment on AWS Lambda, with historical measurements of cold starts, inference speed, and cost.
Run Llama 2 with llama.cpp on AWS Fargate using Docker and AWS Copilot.
Deploy language-model inference on AWS Fargate with Hugging Face Text Generation Inference and AWS Copilot.
Build an ONNX inference endpoint with aiohttp, Docker, AWS Fargate, and AWS Copilot.
Use AWS Lambda SnapStart to reduce initialization overhead for Java-based machine learning inference.
AWS re:Invent 2022 sessions on no-code and low-code machine learning services and their applications.
Twelve AWS re:Invent 2022 sessions on training, deployment, observability, and operating machine learning systems.
Practical considerations for deploying computer vision at the edge with the AWS Panorama appliance.
Seven AWS re:Invent 2021 sessions on deploying and operating machine learning workloads.
Compare training and inference on x86 and AWS Graviton2 Lambda functions across different feature counts.
AWS Summit 2021 sessions on machine learning infrastructure, training, inference, and operations.
A selection of AWS re:Invent 2020 sessions covering machine learning workflows and MLOps.
Prepare satellite imagery, label agricultural fields, and train an Amazon Rekognition Custom Labels model.
Build a training workflow with a custom Docker image, Amazon SageMaker, AWS Lambda, and Step Functions.
Profile a Java application on AWS Batch with Amazon CodeGuru Profiler and orchestrate jobs using Step Functions.
An archived discussion of the advantages and practical challenges of deploying deep learning with serverless infrastructure.
Развёртывание TensorFlow на AWS Lambda: архитектура, подготовка библиотек и модели, API Gateway и Serverless Framework.
An archived attendee guide to seventeen AWS re:Invent 2023 sessions on generative AI and large language models.