Amazon SageMaker AI

2026-11-02

CODING · 1 SOURCE · AWS Machine Learning

Amazon SageMaker AI guide: sequence-level knowledge distillation for efficient agents

Amazon’s SageMaker AI blog post explains how to use sequence-level knowledge distillation (SeqKD) to train smaller, cheaper ‘student’ models that reproduce the multi-step behavior of larger frontier ‘teacher’ models. The post covers generating high-quality agentic trajectories via a teacher API, filtering solved runs with a programmatic scorer, encoding them for supervised fine-tuning, running serverless training, and evaluating the student on a held-out benchmark.

6.0

2026-10-28

CODING · 1 SOURCE · Cohere

Build a citation-backed financial research assistant using Cohere and Anthropic on AWS

A CloudFormation-based demo shows how to build a financial research assistant on AWS that combines Cohere Embed, Rerank, and Parse with Amazon SageMaker AI, OpenSearch Serverless, and Amazon Bedrock (Anthropic Claude for answers). The guide explains using Embed Pro/Fast and Rerank Pro/Fast, Context Focus, and Parse to index page-level content (including tables and scanned pages) and produce citation-backed answers while trading off retrieval quality, latency, and cost.

5.0

2026-10-27

COMPANIES · 1 SOURCE · AWS Machine Learning

Exowatt uses SageMaker serverless customization to fine-tune Qwen3 models for edge and cloud, beating frontier models

Exowatt built a repeatable pipeline on Amazon SageMaker AI serverless model customization to produce task-specific fine-tuned models for edge alarm triage (Qwen3-1.7B) and cloud energy-market analysis (Qwen3-14B). The tuned models outperformed benchmark frontier models on accuracy for these tasks and reached up to 49× lower cost, while enabling a repeatable path from raw operational data to deployable artifacts in days rather than months.

6.0

2026-10-23

COMPANIES · 1 SOURCE · AWS Machine Learning

Hands-on guide: Multi-Turn Reinforcement Learning (MTRL) on Amazon SageMaker AI with AgentCore

Amazon published a hands-on guide demonstrating Multi-Turn Reinforcement Learning (MTRL) on SageMaker AI that uses an agent hosted on Amazon Bedrock AgentCore (or on user infrastructure) to sample policy models and return episode-level rewards. The post walks an end-to-end pipeline (deployment, dataset prep, LoRA training with GRPO, MLflow monitoring, evaluation), uses Wordle as a toy example, reports improving GPT-OSS-20B solve rate from 61% to 84% (and from 15% to 69% at a lower reasoning effort) for about $339 in training tokens, and provides code in an aws-samples repo.

6.0

2026-10-22

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS launches SageMaker AI serverless 'Training Sessions' (GA) for model customization

Amazon announced general availability of Training Sessions for SageMaker AI serverless model customization, letting users run custom post-training loops (SFT, DPO, RLAIF, multi-turn RL, LoRA or full fine-tuning) without managing GPU infrastructure; SageMaker handles GPU allocation, distributed execution, checkpointing and fault recovery and offers a cookbook and agent skills for end-to-end workflows. Customers cited use cases including healthcare claim-denial triage and multi-turn agent training with tools like Amazon Neptune and gpt-oss-20b.

7.0

2026-10-22

CODING · 1 SOURCE · AWS Machine Learning

Measured autoscaling for TTS on Amazon SageMaker using Kokoro-82M

A how-to demonstrates deploying Kokoro-82M (an 82M-parameter non-autoregressive TTS) as a custom container on an Amazon SageMaker AI real-time endpoint via AWS CDK and replacing a guessed autoscaling target with one derived from two live measurements: Q_max (max per-instance concurrency inside the SLO) and T_total (end-to-end scale-out time). The deployment uses SageMaker's native ConcurrentRequestsPerModel metric (reported to CloudWatch) and supports both InvokeEndpointWithResponseStream and InvokeEndpointWithBidirectionalStream transports; tooling converts client-measured concurrency into CloudWatch units and the example run reported a 1.0 ratio between measured and published values.

6.0

2026-10-19

MODELS · 1 SOURCE · AWS Machine Learning

Deepgram ports Nova-3 speech-to-text to AWS Trainium2, gains 11–70% batch cost efficiency

Deepgram worked with AWS to port its Nova-3 speech-to-text model to AWS Trainium2, using TorchNeuron, NKILib kernels and custom Neuron kernels; the final setup runs eight model replicas on a single Amazon EC2 trn2.3xlarge instance. On Deepgram’s offline, throughput-oriented batch workloads (e.g., contact-center analytics, medical transcription) this configuration processed 11–70% more audio per dollar versus the compute options Deepgram previously used, measured using instance pricing at the time of evaluation.

7.0

2026-10-15

MODELS · 2 SOURCES · AWS Machine Learning · The Verge AI

Meta’s Muse Spark Open (816B MoE) added to Amazon SageMaker JumpStart

Meta’s coding-optimized Muse Spark Open (Mixture-of-Experts, 816 billion parameters with 42B active per token and a 1 million-token context window) is now available in Amazon SageMaker JumpStart as two variants (Muse-Spark-Open-816B-A42B BF16 and Muse-Spark-Open-816B-A42B-NVFP4 NVFP4). The model’s open weights (released under the Muse Model License) can be deployed to SageMaker AI endpoints in your AWS account, with instance recommendations ml.p6-b300.48xlarge for BF16 and ml.p5e.48xlarge for NVFP4 and support for the model’s agentic reasoning and multimodal inputs.

8.0

2026-10-14

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon SageMaker JumpStart adds Model Hub browsing, preconfigured endpoint options, and one‑click Hugging Face deploy (2026)

A new AWS blog post explains how to deploy and benchmark pretrained models using Amazon SageMaker JumpStart and a reusable SageMaker Python SDK v3 script that measures latency, output length, and single-request generation rate. In 2026 SageMaker JumpStart added a Model Hub view in the SageMaker AI console, named preconfigured endpoint options (including Neuron‑optimized and cost/latency profiles), and a deep link to deploy Hugging Face models to SageMaker in one click.

5.0

2026-10-12

CODING · 1 SOURCE · AWS Machine Learning

Deploy llama.cpp decision models on Amazon SageMaker AI with AWS DLC and SemIf

AWS published a how-to showing how to use the AWS Deep Learning Container (DLC) for llama.cpp to serve OpenAI‑compatible chat completions via SageMaker InvokeEndpoint, and how to extend that image with SemIf (formerly OpenJev) to return option probabilities and logits for structured decision tasks. The post includes CPU and GPU deployment options, a child‑image build path using CodeBuild/ECR for SemIf, and a benchmark comparing CPU vs GPU latency for SemIf workloads.

6.0

2026-10-02

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon SageMaker AI demonstrates multi-turn reinforcement learning (MTRL) fine-tuning for search agents with Qwen3.6-27B

Amazon’s blog post describes using Amazon SageMaker AI’s multi-turn reinforcement learning (MTRL) capability to fine-tune a Qwen3.6-27B model for an LLM-powered search agent. The write-up details MTRL features—serverless execution, modular agent-environment interfaces, asynchronous rollouts, built-in algorithms (PPO, CISPO, IS), trajectory observability via MLflow, and evaluation jobs—and shows an enterprise search use case combining BM25 and vector search tools.

6.0

2026-10-01

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS documents Amazon Bedrock AgentCore multi-agent workflows and integrations for cloud migrations and agent workloads

An AWS technical post explains how AWS Professional Services uses a multi-agent framework built on Amazon Bedrock AgentCore to automate end-to-end enterprise cloud migrations and reduce infrastructure-as-code development time from weeks to minutes. The post also covers related how-tos and integrations — Live Data in Amazon QuickSight, using Amazon S3 Vectors as persistent memory with NVIDIA NeMo Agent Toolkit on EKS, ambient/ambient-agent patterns, Claude Platform configuration on AWS, personalization with SageMaker AI, fine-tuning Nova 2 Lite for uniopen, a Bedrock Knowledge Bases conversational claims assistant, and AgentCore Runtime Instances for colocated GPU agent workflows.

7.0

2026-10-01

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon Payments deploys LinUCB contextual bandit on Amazon SageMaker AI to personalize acquisition funnel

Amazon Payments implemented a multi-objective contextual multi-armed bandit (LinUCB) on Amazon SageMaker AI to personalize content across a three-step acquisition funnel. In a seven-week online A/B test they report a high single-digit percentage relative lift in final-funnel conversion for one customer population while another saw no improvement; the team published the AWS architecture, method details, and a code repository for experimentation on synthetic data.

6.0

2026-10-01

COMPANIES · 1 SOURCE · AWS Machine Learning

Uniopen fine-tunes Amazon Nova 2 Lite for retail content moderation

Uniopen (part of Taiwan’s Uni‑President Enterprises Group) adapted Amazon Nova 2 Lite for its retail moderation taxonomy by supervised fine-tuning in Amazon SageMaker AI and additional prompt‑level output optimization, using Nova 2 Pro to generate candidate corrections that humans verify before inclusion in training data. The deployment pipeline uses Amazon S3, DynamoDB, Argo Workflows on EKS, Argo CD, SNS, CloudWatch and Bedrock Guardrails with hard/soft gate checks, fixed test sets, and mandatory human review for ambiguous cases.

6.0

2026-09-28

CODING · 1 SOURCE · AWS Machine Learning

Deploy streamed Qwen3-TTS on SageMaker using vLLM-Omni v1.5 for real-time voice

AWS published a Part 1 tutorial showing how to deploy the Qwen3-TTS text-to-speech model on Amazon SageMaker AI using the vLLM-Omni v1.5 Deep Learning Container (DLC). The guide demonstrates sending text and receiving audio chunks over a single persistent bidirectional connection (SageMaker bidirectional streaming) via the vLLM-Omni native WebSocket route (v1/audio/speech/stream) and includes a Gradio example; the series will also cover other specialized DLCs like WhisperX and llama.cpp.

6.0

2026-09-28

CODING · 1 SOURCE · AWS Machine Learning

Generate images and image-conditioned video on SageMaker AI using AWS vLLM-Omni DLC (FLUX.2-klein-4B → Wan2.1-VACE-1.3B)

An AWS blog post demonstrates deploying the same AWS vLLM-Omni Deep Learning Container to two SageMaker AI endpoints: a real-time endpoint running FLUX.2-klein-4B for image generation and an asynchronous endpoint running Wan2.1-VACE-1.3B for image-conditioned video generation. The sample workflow generates a PNG from text, passes it (via S3) to the video endpoint, retrieves the MP4 from S3, and includes CLI and optional Streamlit examples.

5.0

2026-09-24

MODELS · 1 SOURCE · AWS Machine Learning

AWS packages WhisperX DLC for SageMaker to deliver speaker‑labeled, per‑word timestamps

AWS provides a WhisperX Deep Learning Container (DLC) for Amazon SageMaker that bundles OpenAI Whisper with wav2vec2 forced alignment and speaker diarization to produce per-word timestamps and speaker labels. The GPU-ready container conforms to SageMaker’s serving contract, supports real-time and asynchronous endpoints, and outputs json, verbose_json, srt, and vtt formats (example image tag: whisperx:3.8.6-cu128-amzn2023-sagemaker).

6.0

2026-09-22

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon SageMaker AI adds built-in concurrency sweeps to right-size generative AI endpoints

Amazon describes using concurrency sweeps—integrated in SageMaker AI Inference Recommendations—to benchmark generative AI endpoints by increasing simultaneous request levels and measuring throughput and latency. The post walks through deploying the NVIDIA Nemotron-3 Nano 30B on an ml.g7e.2xlarge with the vLLM container and running CreateAIBenchmarkJob (with workload defined via CreateAIWorkloadConfig) to discover saturation points and right-size capacity.

6.0

2026-09-21

CODING · 1 SOURCE · AWS Machine Learning

Posit's Positron IDE runs on Amazon SageMaker AI with Bedrock-enabled Posit Assistant

Posit published a Positron container image definition that can be registered with Amazon SageMaker AI so administrators can attach the IDE to a SageMaker Studio domain. Running Positron on SageMaker AI lets data scientists query governed AWS data (Athena, Glue, S3) under a Space execution role, launch Spaces on reserved or on-demand SageMaker compute, collaborate in shared or parallel Spaces, and use Posit Assistant with Amazon Bedrock as the model provider inside the customer's AWS account and region.

6.0

2026-09-21

COMPANIES · 1 SOURCE · AWS Machine Learning

EXL launches AI-powered Medical IDP solution on AWS

EXL has combined its Xtrakto.AI intelligent document processing application with a domain-tuned EXL Insurance LLM and deployed the pipeline on AWS (using Amazon SageMaker AI and Amazon Bedrock among other services) to extract, summarize, and query medical records at enterprise scale, aiming to cut claims and underwriting review time. The end-to-end solution runs within an AWS Region and uses AWS services for ingestion, orchestration, OCR, model training/inference, and secure access control.

5.0

2026-09-18

COMPANIES · 1 SOURCE · AWS Machine Learning

AWS shows how to migrate multi-model Hugging Face smolagents to Amazon Bedrock AgentCore runtime

AWS published a how-to that migrates a multi-model healthcare AI agent built with Hugging Face smolagents from a self-managed Amazon ECS/Fargate deployment to the managed Amazon Bedrock AgentCore runtime. The solution preserves existing agent logic while offloading container lifecycle, scaling, identity, and observability to AgentCore and demonstrates orchestration across Llama 3.1 70B Instruct on Bedrock, BioM-ELECTRA-Large-SQuAD2 on SageMaker (and containerized), plus vector retrieval via Amazon OpenSearch Service.

6.0

2026-09-18

CODING · 1 SOURCE · AWS Machine Learning

Hugging Face publishes six open-source Skills to deploy models on Amazon SageMaker AI via coding agents

Hugging Face published six open-source 'Skills' (GitHub) that let coding agents orchestrate end-to-end deployment of Hugging Face models to Amazon SageMaker AI. The skills automate selecting the correct serving container from AWS Deep Learning Containers, creating real-time or serverless endpoints with autoscaling and CloudWatch alarms, and provide verified teardown paths; they run using Python and the AWS CLI and are designed to prevent fragile or costly agent-made deployment mistakes.

6.0

2026-09-17

MODELS · 1 SOURCE · AWS Machine Learning

Amazon SageMaker deploys Qwen-Image-Edit-2509 and Rekognition to augment industrial-safety datasets with synthetic people

Amazon demonstrates an end-to-end pipeline on SageMaker AI that uses the diffusion model Qwen-Image-Edit-2509 (hosted on an ml.g5.12xlarge with NVIDIA A10G GPUs) to insert photo-realistic synthetic people into real equipment images, then generates bounding-box annotations automatically via Amazon Rekognition DetectLabels; experiments reportedly showed up to 160% improvement in person-detection mAP50 without manual annotation. The workflow edits real scenes in-place to preserve background fidelity, deduplicates Rekognition boxes with NMS, and converts labels to YOLO format for training.

6.0

2026-09-15

COMPANIES · 1 SOURCE · AWS Machine Learning

Amazon SageMaker AI adds instance preference lists for training and processing jobs

Amazon SageMaker AI now supports instance preference lists for training and processing jobs. You can specify an ordered list of up to five instance types, and SageMaker AI will automatically launch the first type with available capacity, removing the need for manual retry loops and capacity‑watching scripts.

6.0