Sydney AI Training

This program is 3 day industry training and hackathon designed for software engineers and technical professionals who wants to build, deploy, and compete with real AI systems. It is suited to:

Software engineers and developers building or integrating AI into their products
Technical professionals working with large language models, agents, and AI infrastructure
Engineers looking to gain hands-on experience with model evaluation, fine-tuning, and agentic frameworks
Developers interested in the latest tooling from NVIDIA, LangChain, and the open-source AI ecosystem
Participants should have a working knowledge of Python and be comfortable in a Linux/command-line environment. Capacity is limited to 40 attendees

Trainers from NVIDIA, Anthropic and LangChain

Day 1: Building with AI

Session 1: Where the Future Economy is Headed with Software 2.0

Duration: 2 hours | Format: Presentation & Discussion

Creating value in the future economy — electricity + compute + developers = revenue
Evolution of software architectures — from monoliths to AI Agents
The death of SaaS and the age of personal software
The path to reasoning — RAG techniques and limitations of LLMs
Agentic applications — what they are and why they matter
The economics of inference — cost per token, latency vs. throughput tradeoffs, why this shapes product decisions
Edge AI and the on-prem trend
Gigabyte AI developer hardware introduction

Session 2: Hands-On Lab — Building with AI, transitioning to no-manual coding

Duration: 6 hours | Format: Lab-based

Dependencies: User to work on own laptop, requires WSL, Ubuntu
Attendees receive a Claude 1 month voucher.

Greenfield Software Development (4 hours)

Prompt Engineering Fundamentals
Building skills, skill composability
Context Engineering and memory (Vector RAG, Graph RAG) - Obsidian & Graphify
Connecting to agents using MCP servers
Planning and designing
Testing & iteration

Legacy Code & Migration (2 hours)

Re-engineering and migrating legacy codebases (clean-room re-implementation)

Patterns & Methods & Workflow

Architectures and patterns for AI-assisted development
Reducing token usage (tied to context engineering and memory)
Ways of working: the AI-augmented developer workflow
Human in the loop, code reviews, and QA
Debugging AI apps in production using LangSmith

Day 2: Building AI — AI Engineering Foundations

Duration: 8 hours

Session 3: Infrastructure & Models

Duration: 3 hours

GB10 (Gigabyte AI Top Atom) clustering
Environment set up, dependencies
Hosting foundational models and short-term memory

Session 4: Evaluation & Light Customization of LLMs

Duration: 2 hours

Based on NVIDIA's Deep Learning Institute courseware ref

LLM Evaluation Fundamentals
Running benchmark evaluation using GSM8K ref
Parameter-Efficient Fine-Tuning (PEFT) with Low-Rank Adaptation (LoRA)
NVIDIA NeMo Framework, NeMo Evaluator, and NeMo Customizer
NVIDIA Inference Microservices (NIMs)
MLflow for Experiment Tracking

Session 5: Agents

Duration: 3 hours

LangChain Foundations (Python)

Building simple agents
Model Context Protocol (MCP) — what it is, building an MCP server, connecting to third-party MCP servers
Multi-agent protocols
Setting up a coding agent

VLMs and VLQA

Openclaw

Openclaw/Nemoclaw setup
Setting your own coding agent (Clawcode)
Claw automation tutorial (e.g. email automation agent)

Day 3: Hackathon