AI RESEARCH AND ITERATION AGENT

CoreWeave
ARIA

Whether you're building an agent or training a model, ARIA (AI Research and Iteration Agent) walks you through the AI Loop. It analyzes your experiments, proposes what to try next, and launches it with your approval, showing its evidence with in-chat visualizations at every step.

Put every experiment to work

Drive continuous improvement

Analyze results, propose the next experiment, launch it, repeat. Set it up once as an Automation and the loop runs on every finished experiment. Connect your codebase and ARIA drafts each change as a PR, so nothing ships without a human reviewing it.

Accelerate deeper analysis

Training models and building agents is data-heavy work, and understanding it takes more than numbers. ARIA delivers live visualizations in the Forge dashboard, so you can follow its reasoning clearly. When a finding matters, it creates persistent reports and dashboards your team can keep using.

Shared team knowledge

Forge is organized around projects, and so is ARIA's memory. As your team works, ARIA remembers what the experiments showed and what you decided, scoped to the project rather than to one person's session. Every conversation starts from shared team learnings.

Autoresearch

Continuous improvement on autopilot

Describe what you want to learn and ARIA forms hypotheses, launches experiments, evaluates results, and recommends the next iteration. For agent builders that means prompt and model changes tested against real traffic. For model trainers it means training jobs launched on your infrastructure, from sweeps to fine-tuning and RL. You approve each launch and review each change on its own branch, while ARIA handles the work in between.

WORKSPACES AND REPORTS

Live, persistent dashboards to back up findings

Every finding comes with the visualization behind it, rendered live in the chat against your project's data. Standard charts cover the common questions; for anything beyond them, ARIA builds custom interactive plots on the fly, from distributions and heatmaps to scaling fits. When an insight is worth keeping, ARIA turns it into a report or a dashboard view that updates as new experiments arrive, so the evidence stays current long after the conversation ends.

CONTEXT

Full experiment context, already loaded

ARIA knows how to query your project's experiment data efficiently, so you jump straight to the question and get answers grounded in the full history of your research. It also knows what you're looking at. Open it from a crashed experiment's log, a table, or a report and the conversation begins there. And you can hand it exactly the context a question needs, adding panels, evaluations, or agent traces to the chat for a focused analysis.

Mobile app

Close your laptop. ARIA keeps going.

ARIA runs in the cloud, not on your machine. Long analyses and training launches keep going after you step away, Automations fire on every finished experiment whether you're watching or not, and the conversation is waiting when you come back, on your desktop or in the mobile app.

"ARIA has become a valuable part of my daily workflow. It helps me quickly generate reports, create sweep configurations from natural language, and automate tasks that would otherwise require a lot of manual setup. Even today, the agent saves time on repetitive work and makes interacting with W&B much more intuitive."
Praneeth Gangavarapu, PhD Candidate
Scripps Research
ARIA IS PART OF COREWEAVE FORGE

ARIA runs at every stage of the AI loop

Built into CoreWeave Forge, ARIA closes the loop between experimentation and improvement, using what it learns from every run to guide the next one.

Run

Observe

Curate

Improve

Evaluate

Related resources

The same news reads differently depending on where you sit. Here’s the version that applies to you.

Blog

Introducing CoreWeave ARIA: AI Research and Iteration Agent

Video

Accelerate the Self-Improving AI loop with CoreWeave ARIA

Blog

CoreWeave Forge: Turn AI Iteration Into Compounding Improvement

Builder Resource Center: Learn from every run

Explore demos, code, and technical resources for every stage of the AI loop. Learn how researchers, developers, and CoreWeave engineers build, observe, evaluate, and improve AI systems—and put those insights to work.