Tag
FlashRT is an agent harness that guides coding agents to automatically optimize and deploy real-time multimodal applications, achieving up to 70x latency reduction on NVIDIA B200 GPUs and 3.6x throughput improvement on AMD MI355X.
Netflix shares the design decisions behind its in-house LLM serving stack, including engine selection (vLLM), model packaging, API surface, and deployment strategy, highlighting trade-offs revealed under production load.
You can build, deploy, and evaluate long-running AI agents using Claude Code, Codex, or Antigravity without writing code.
Hoplite is a product that enables effortless deployment of cloud software factories, as featured on Product Hunt.
The article argues that the hardest part of deploying a private AI rollout is not setting up the VPC, but building custom connectors for internal tools that lack off-the-shelf support, a factor often underestimated in procurement.
A detailed guide on running the quantized NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B model on two RTX 3090s using vLLM with full 262K context, achieving high inference speeds without CPU offloading.
This paper introduces safeguard-conditioned uplift, a protocol for measuring how different deployment access conditions (e.g., helpful prompting, safety prompting, external safeguards) affect the utility-risk frontier in dual-use biology AI assistants, based on human-judged evaluations of Claude Sonnet 4.6 and Gemini 3.5 Flash.
This paper formulates the Foundation Model Deployment Portfolio problem to minimize total cost of ownership for deploying LLMs and VLMs across transportation management functions, considering quality, latency, and safety constraints. A case study shows cost savings of 97% compared to all-closed-API baselines.
SPINE is an agentic framework that systematically debugs and deploys bimanual robots, reducing reliance on expert calibration and improving operationalization success across platforms.
The article argues that the best AI model is not necessarily the most powerful, but the one that can be practically deployed and run efficiently, emphasizing the importance of considering real-world constraints like cost and hardware requirements.
The article discusses the significant challenges enterprises face when deploying AI agents to production, highlighting the lack of standard deployment infrastructure, security concerns, and the need for an orchestration layer to manage agent lifecycles.
A library of 150+ AI use cases documenting who is deploying AI and the outcomes.
An analysis of the key factors that will differentiate production-ready AI agents from demo prototypes in the coming years, focusing on reliability, scalability, and real-world integration.
Google AI Studio now supports direct app publishing and gives you a free subdomain. No need to buy a domain or configure servers. Accessible globally in minutes.
ManusAI introduces auto-publish, which automatically deploys successful builds to the live URL with off-by-default setting and queuing capability, removing the last manual step between building and shipping.
Unsloth collaborated with AWS on a comprehensive guide to dynamic quantization and deployment of LLMs on Amazon SageMaker, covering model formats, tools, and best practices.
ByteDance's Seed team has released the Seed2.0 model card, detailing a model designed to bridge the gap between lab benchmarks and real-world software engineering. The card highlights deployment tiers, performance comparisons, and honest acknowledgment of gaps versus frontier models.
Ant is a JavaScript runtime and ecosystem including an engine, package manager, registry (ants.land), deployment platform, and desktop app framework (Ant Desktop) designed as a coherent alternative to existing JavaScript stacks.
A curated list of open source libraries for deploying, monitoring, versioning, scaling, and securing production machine learning systems.
The author questions whether deploying and scaling AI agents for production is a universally frustrating problem, citing issues like hallucinations and state management.