![\"Rio](\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2026\/05\/1779763993_905_ImageResizer.ashx.png\")
<\/p>\nRio Tinto has built an AI domain assistant that documents the knowledge, dependencies and decision logic used by a key manufacturing system in its aluminium operations.<\/p>\n
<\/p>\n
\n Rio Tinto’s Ke Shi.
\n <\/p>\n
Data science senior advisor Ke Shi told the AWS Summit Sydney that the assistant is unlocking \u201caccumulated knowledge and complexity\u201d embedded in one of its most critical systems, a 30-year-old manufacturing execution system (MES) called Metpro.<\/p>\n
Metpro, he said, manages the aluminium product lifecycle end-to-end, \u201cfrom tapping right through to shipments, and acts as a central hub connecting the process control system, data capture platform and operational infrastructure.\u201d<\/p>\n
Shi said that over time, Metpro\u2019s technical documentation had become fragmented, spread across \u201cthousands\u201d of documents.\u00a0<\/p>\n
Additionally, the system\u2019s highly coupled nature meant that dependencies were not always documented or well understood.<\/p>\n
\u201cThat means even a small change like a UI configuration can have very unexpected downstream impact,\u201d Shi said.<\/p>\n
\u201c[Altogether] for engineers, this translates into slow onboarding, difficulty in finding the right expertise, and high risk when making changes.\u201d<\/p>\n
Rather than try to rewrite or replace Metpro, given how embedded it is in daily functioning of Rio Tinto Aluminium Pacific Operations, Shi said the company instead looked for ways to \u201cretain and operationalise decades of embedded knowledge.\u201d<\/p>\n
\u201cWe needed a way to preserve how the system actually works, including dependencies and the decision logic that were incomplete or had become outdated over time, and make that accessible to engineers without changing the underlying platform,\u201d Shi said.<\/p>\n
\u201cOur goal was to bring together the fragmented technical documentation and turn them into something more coherent, rich and usable.\u201d<\/p>\n
The miner started by building a \u201cdomain-aligned training dataset\u201d, plugging the Metpro code base and \u201cbusiness and operational context\u201d into Amazon Bedrock Knowledge Bases and Amazon Bedrock AgentCore.<\/p>\n
The company then used Amazon SageMaker AI Jumpstart – a service used to access, evaluate and compare foundation models – and adopted Llama 3.1 8B as the initial inference model.<\/p>\n
\u201cUsing our curated dataset and Amazon SageMaker AI Jumpstart, we trained the model to internalise how the system actually behaves and more interestingly, how we expect it to respond,\u201d Shi said.<\/p>\n
\u201cFine-tuning gave the model a strong domain understanding, but the real world doesn\u2019t stand still. Knowledge changes, processes evolve, and new information appears every single day.\u00a0<\/p>\n
\u201cAgents help bridge that gap by working alongside the trusted knowledge, pulling up to date information as needed, without having to retrain the model every single time.<\/p>\n
\u201cThe agent layer keeps the understanding [of the Metpro system] refreshed, reliable and connected to reality.\u201d<\/p>\n
![\"\"\/](\"https:\/\/i.nextmedia.com.au\/News\/Metpro.png\")
<\/a><\/p>\n