Data Science Dojo Singapore

As enterprises move from simple prompts to complex, autonomous workflows, the challenge shifts to effectively integrating multiple components—such as different LLMs, external tools, and memory—into reliable, production-grade Agentic Applications.
In this session, we’ll introduce FloTorch, a framework designed to simplify the development, management, and deployment of these complex AI agents. FloTorch provides the necessary structure and capabilities to integrate, observe, and secure agents across different model providers.
We’ll begin with an overview of the core architectural challenges in building sophisticated agents-from managing conversational memory and orchestrating tool use, to ensuring reliability and performance across multiple LLM calls.
Then, we’ll show how FloTorch provides a unified workflow for integrating different LLMs, various memory stores, and custom tools, helping teams prototype, benchmark, and operationalize Agentic Applications without technical debt. We will also highlight its built-in Observability features, including Traces and OpenTelemetry integration (where available), which are critical for debugging and monitoring agent behavior in production.
A live demo will walk through building a scalable agent with FloTorch, highlighting how developers can add memory, define tool-use logic, and integrate crucial observability into their pipelines.

### What We Will Cover:

• Key challenges in architecting and deploying autonomous AI Agents at scale
• How FloTorch enables seamless integration of different LLMs, Memory, and Tools
• Designing robust, multi-step agentic workflows for business use cases
• Implementing Observability (Traces, OpenTelemetry) to monitor and debug agent decisions
• Strategies for ensuring safety, reliability, and governance in agentic applications
• Real-world examples of business value driven by FloTorch-powered agents

### Hands-On Insights:

Through live demonstrations and interactive Q&A, participants will gain hands-on insights into using FloTorch to rapidly build, evaluate, and operationalize complex Agentic Applications—with integrated tools for memory, LLM switching, and production-ready observability.