The Hallway Track
16 tracked signals on MLOps.
LangChain launches Engine, an AI that autonomously improves other agents through continuous development cycles.
“They've established a continuous agent development lifecycle. Build, test, deploy, monitor, and then repeat, so that they can learn from real usage and iterate quickly.”
AI agents lack persistent state checkpoints, making debugging and replay impossible today.
“all of that is lost and it is only stamped as a read-only trace by the end, which is sitting in another tool far away from where the actual code is.”
NVIDIA introduces Green Contexts for fine-grained GPU resource sharing between concurrent workloads
“Controlling how GPU resources are shared between them remains difficult.”
NVIDIA ModelExpress accelerates multi-hundred-GB model weight distribution across GPU clusters
“Every byte moved has a cost. As model checkpoints grow to hundreds of gigabytes or even a terabyte, that cost adds up quickly.”
Lyft applies ML model evaluation rigor to AI agents before production deployment.
“if we are running um offline evaluations for our machine learning model before that goes to productions, I think we should do the same for AI agents as well”
Microsoft Foundry positions model operations, not model access, as the core AI challenge
“The hardest part of building AI systems today is no longer getting access to a capable model. It is knowing how to choose, validate, optimize, and operate the right model across the full lifecycle of a real application.”
NVIDIA releases AICR v1.0 for open, verifiable GPU cluster configuration on Kubernetes
AWS adds Studio UI to manage HyperPod Spaces, eliminating CLI dependency for ML developers.
“reducing the time from cluster access to productive development to a few clicks”
LangChain defines a four-phase agent development lifecycle: build, test, deploy, monitor.
“What is the agent development lifecycle? So the agent development lifecycle is the process that agent engineers use to build and improve their agents.”
Quality beats quantity in SFT: 1,000 curated examples can match models trained on far more data.
“A wrong demonstration gets imitated.”
AWS outlines four-layer governance model for shared SageMaker HyperPod ML compute clusters
Deepgram adds billing transparency and GPU-level observability to SageMaker AI speech deployments.
Jumio built a sub-100ms real-time feature store on AWS for fraud detection ML models
NVIDIA outlines full-stack observability strategy for multi-layer AI factory infrastructure
SageMaker Python SDK v3 now exposes AI inference recommendations directly in notebook workflows
AWS introduces inference meta-monitoring for SageMaker endpoints to continuously track ML model prediction quality in production