From post-training to production: How TinyFish runs its web-native model on Databricks AI Runtime
TinyFish runs web-native agents on Databricks AI Runtime—a production case study in agentic infrastructure.
Why it matters
A concrete deployment showing how enterprises move agents from model-serving to live-web autonomy. The story illustrates the integration layer (Databricks runtime + web-agent framework) and operational requirements for agents that fetch and act across the live web.
The key facts
12 to knowTinyFish builds agents that operate across live web (Search, Fetch, Browser capabilities)
Deployment uses Databricks AI Runtime as the inference/operational foundation
Post-training to production pipeline documented with real engineering detail
Agent type: web-native autonomy (multi-step retrieval and action)
Infrastructure choice: Databricks (not generic cloud, specific platform)
No revenue, customer count, or deployment scale disclosed
Source: vendor blog (Databricks); no independent validation
TinyFish builds agent infrastructure for enterprise web automation
Uses Databricks AI Runtime for model serving and post-training
Web-native model design: agents that operate across live web contexts
Case study format: no independent validation or comparative benchmarks provided
Published by Databricks (vendor blog)
Go to the source
Databricksdatabricks.com
Publisher excerpt: TinyFish builds infrastructure for enterprises that need agents to work across the live web. Its Search, Fetch, Browser...