Fall Launch 2026: AI to Build Agents and Models
Watch the Fall Launch and explore how FAiE brings agent development, evaluation, data, training, and deployment into one workflow.
Here's what we launched
Build an agent with FAiE
Manos
See the workflow for building a support agent, comparing models, and improving it with production feedback.
- Connect tools and evaluate agent behavior
- Train, deploy, and monitor the underlying model
What's new in Oumi
Mary
Mary introduces agent development, Bring Your Own Harness, and Direct Preference Optimization.
- Evaluate prompts, tools, and models together
- Improve your agent while keeping your existing infrastructure
Agent development and Bring Your Own Harness
Mary and Matthew
Build an agent end to end or connect an existing system to Oumi for evaluation and improvement.
- Generate multi-turn agent data and test tool use
- Turn failure patterns into targeted training data
Train with preference feedback
Mary and Jeremy
Learn when to use supervised fine-tuning and when preference pairs can refine agent behavior.
- Establish behavior with supervised examples
- Use DPO to refine policy adherence and communication
Build and own specialized intelligence
Manos
Manos closes the launch with the case for building intelligence around your business.
- Evaluate and improve complete agents
- Keep building on the intelligence you own
Agents for financial services
Damon
Explore agent workflows for financial services, from policy lookup to human handoffs.
- Evaluate tool use and policy adherence across interactions
- Use production feedback to inform the next approved release
