AI Talk: Operationalizing Reinforcement Learning for Customer Support AI (подія в архіві)

For this session, we invited experts from Weights & Biases and OpenPipe to demonstrate how reinforcement learning can be used to train support agents to make more effective decisions.
Instead of relying on labeled “correct” answers, the agent explores different actions, such as resolving a ticket or escalating, and learns from structured reward signals provided by an automated judge model.
We’ll walk through the full training pipeline, including synthetic ticket generation, adversarial scenario design, and evaluation workflows.
You’ll see how reinforcement learning algorithms can be applied to optimize triage policies, how reward models shape agent behavior, and how training stability is managed through monitoring techniques.

Speakers:
Jorge Silva, AI Solution Engineer at Weights & Biases. 7+ years of experience building and deploying production AI and developer tooling across enterprise and startup environments, from embedded systems and assistive technology to full-stack ML applications and customer-facing AI platforms. Background in solutions engineering, ML observability, developer infrastructure, and end-to-end delivery from prototype to production.
Daniel Bolus, Senior Product Manager and Solutions Architect at OpenPipe. 6+ years of experience leading product innovation, startup operations, and community-driven tech initiatives across health tech, hardware, and digital marketplaces. Background in biomedical engineering, startup leadership, and end-to-end product ownership from idea to launch.

🗓 April 16, 5:00 PM EEST
📍 Online
🗣 In English
REGISTER hubs.ly/Q049_0VX0

💡 Усі статті, обговорення, новини про AI — в одному місці. Приєднуйтесь до AI спільноти!