Join us for another insightful and hands-on meetup where we dig into what it actually takes to turn an LLM into a domain-specific agentic tool that a real business relies on. Sizing picking the right pump, tank, or equipment spec for a customer's exact requirements is normally a slow, expertise-heavy process. In this session, we'll walk through how the OpenAI Agentic SDK was used to build an agentic sizing tool at Davis & Shirtliff: an agent that reasons through technical requirements, calls the right tools, and returns a sized recommendation the way a human specialist would.

Whether you're a beginner curious about what "agentic" actually means beyond the buzzword, or an experienced engineer looking to see a real production use case of the OpenAI Agentic SDK, this session is for you. Each part of the build comes with a live look at the tool in action: the reasoning trace, the tool calls, and the final sized output.

Agenda

Here's our agenda for the day:

1. Welcome and Introduction (15 minutes)
Meet the community and organizers.
Overview of today's theme: "Building an Agentic Sizing Tool with the OpenAI Agentic SDK."
Why sizing is a hard problem for LLMs to solve naively, and why an agentic approach not just a single prompt is what makes it work.

2. Agent Design & Tooling (35 minutes)
Anatomy of the Agent: How the OpenAI Agentic SDK structures agents, tools, and handoffs and how that maps onto a real sizing workflow (requirements gathering → calculation → recommendation).
Tool Definitions for Domain Logic: Wrapping Davis & Shirtliff's sizing calculations and product catalog as callable tools the agent can reliably invoke. Live demo: the agent walking through a real sizing request end-to-end.

3. Handling Ambiguity & Guardrails (40 minutes)
Ambiguous Customer Inputs: What happens when the customer doesn't give the agent enough information clarifying questions, fallback logic, and keeping the agent from guessing on technical specs. Live demo: an underspecified request and how the agent handles it.
Guardrails & Validation: Making sure the agent's recommendation is technically sound before it reaches a customer validation layers, confidence thresholds, and when to escalate to a human.
Putting it together: Lessons learned from taking this from a prototype to something used in a real business context.

4. Q&A and Community Showcase (10 minutes)
Ask questions, share your work, and get feedback from the community.

Gigs

We would love to reach out to you so that you can build for our customers, please fill out this form with details to ensure we have your details:
GIG/HACK DEVELOPER PORTFOLIO FORM

Join community channels

Africa's Talking AI/ML Community:

• Join the Africa's Talking AI/ML Community here

Slack
Join Africa's Talking slack community here

Please follow our Twitter handles to:

• @AfricasTalking

• @ElarianHQ

You can get our videos, recaps, and event interviews on our youtube channels, subscribe to get updates:

• Africa's Talking Youtube Channel

Africa's Talking community allows developers to learn skills for the modern-day African Developer. We are language and framework agnostic. All developers are welcome. This is where Africa's Talking developers community meets to build, learn and exchange knowledge.

We are helping software developers and businesses to bring their ideas to life through easy-to-use APIs easily.

Would you like to partner with us? Kindly contact the Developer Experience Team.