Skip to content

Details

Hey everyone, join us for the August edition of the Agentic Engineering Meetup! We'll have two exciting talks, great discussions, and plenty of opportunities to connect with fellow engineers, AI practitioners, and technology enthusiasts.

🎉 Seats are limited, so RSVP early to secure your spot.

🏠 Location
Nimbly
Darling Park, Tower 1, 201 Sussex St
Sydney NSW 2000, Australia

🍕 Food & Refreshments
Kindly provided by Nimbly😊

Schedule
🕠 5:30 PM – Doors Open, Refreshments & Networking
🕕 5:45 PM – Welcome & Announcements
🎤 5:50 PM – Talk 1: Objective Truths Over Oversight: How We Let AI Agents Ship Correct Code.
🎤 6:05 PM – Talk 2: The Token Tax on Autonomy: why agentic coding is a capacity problem, not a licence.
🎤 6:40 PM – Talk 3: Plug-In, Optimise, Ship: Scaling Open-Source Agent Infrastructure with NeMo
🤝 7:10 PM – Networking
🚪 7:30 PM – Event Close

Speakers:
-----------------------------------------------------------------------------------
🎤 CHRIS BALLMENT, Co-Founder & Chief Architect, Nimbly
Talk Title: Objective Truths Over Oversight: How We Let AI Agents Ship Correct Code.
Talk Summary: The instinct when handing work to an AI coding agent is to watch it closely - to review every step and pull it up the moment it drifts. That doesn't scale, and it wastes the very thing that makes agents useful: their autonomy. This talk makes the case for the opposite approach. Instead of oversight, we give agents agency - bounded by objective truths they can't argue with.
-----------------------------------------------------------------------------------
🎤 CHRISTOPHER SIMUSOKWE, Head of Data Engineering,
TPG Telecom
Talk Title: The Token Tax on Autonomy: why agentic coding is a capacity problem, not a licence.
Talk Summary: Most teams still price AI-assisted coding the way they price software: seat price times headcount, book the number, move on. That worked when the assistant autocompleted a line at a time. It does not survive contact with agents.
An agentic coding task is not one request. It is a loop. The agent reads files, reasons, proposes an edit, re-reads, runs something, sees an error, tries again, and every turn re-sends a growing slice of context. That loop is what you are actually buying, and it behaves like compute, not like per-seat SaaS.
Here is the part that gets missed. Two teams on the same model, running the same workload, routinely see bills an order of magnitude apart. The gap is not commercial. It sits in retrieval, tool design, and how much context you carry per turn. The cheapest token is the one you never send, and deciding which tokens you never send is an architecture decision.
What you will leave with:
- A working mental model for where tokens actually go inside an agent loop
- The four cost drivers, and which ones you control in architecture rather than in billing
- The levers that close the 1x to 10x gap, and the order to apply them in.
-----------------------------------------------------------------------------------
🎤 MARKO SLABAK, Senior Solutions Architect, NVIDIA
Talk Title: Plug-In, Optimise, Ship: Scaling Open-Source Agent Infrastructure with NeMo
Talk Summary: Building agents is getting easier. Running them reliably, efficiently, and securely in production is less so. In this session we’ll look at how two open source projects from NVIDIA tackle that challenge:

- NeMo Agent Toolkit for profiling, evaluation, and multi-agent orchestration across any framework,
- NemoClaw for zero-trust agent runtimes with security baked in. No framework migration required, no vendor lock-in.
We’ll close with a walkthrough of the NemoClaw for LangChain Deep Agents Code blueprint available on build.nvidia.com. This blueprint serves as a starting point to spin up and customise open-model coding agents whether cloud-hosted or on your own hardware.
-----------------------------------------------------------------------------------

We look forward to seeing you there!

You may also like