What I built, how I approached it, and what I learned.

RAMIFY OS — building an auditable product-trust demonstration

RAMIFY OS is a university industry project focused on product trust and auditable AI decision support. As Project Manager, my role sits across requirements, team coordination, testing, documentation, demonstrations and client feedback.

My contribution

• Coordinated scope, responsibilities, sprint priorities and client-facing updates.
• Translated feedback into practical work items and helped keep implementation aligned with requirements.
• Supported functional testing, documentation and final demonstration preparation.
• Worked around the idea of decision receipts and traceable evidence rather than treating AI output as a black box.

What I learned

Good technical delivery is not only about writing code. Requirements, test evidence, clear communication and traceability matter when a system needs to be trusted by other people.

TasNetworks × UTAS × AWS Open Innovation Hackathon

I took part in a three-day open innovation hackathon built around a real Tasmanian energy-sector challenge. The experience combined rapid problem definition, teamwork, idea development and practical AI/cloud thinking under a short deadline.

How I approached it

• Worked with a team to understand the problem before jumping directly into a technical solution.
• Considered how data, AI and cloud services could support a practical outcome.
• Contributed to rapid iteration, communication and presentation of the team's direction.

What I learned

Short innovation projects reward clarity and prioritisation. A smaller solution that clearly addresses the user's problem can be stronger than a technically ambitious idea that cannot be explained or delivered.

Machine Learning Model Comparison — choosing models with evidence

This project compares supervised classification approaches using Python and scikit-learn. The emphasis is not only on training models, but on evaluating them consistently and understanding why one model may be more suitable than another.

Technical approach

• Prepared structured data for classification workflows.
• Compared multiple classification models using consistent evaluation methods.
• Used metrics such as accuracy, precision, recall and F1 alongside cross-validation and tuning.
• Looked at trade-offs rather than treating a single score as the whole story.

Cloud Data Processing — preparing data inside a distributed workflow

This project involved building and testing a transformation worker as part of a distributed Nectar Cloud workflow. The worker prepared structured data for downstream machine-learning use.

Technical approach

• Processed and transformed a prepared dataset containing 1,221 records.
• Encoded categorical variables and prepared numerical features.
• Worked with a distributed workflow where separate services handled different stages of the data pipeline.
• Tested the transformation stage so the output remained predictable for downstream processing.

What I learned

Data engineering becomes easier to reason about when each stage has a clear responsibility, defined input/output and tests that verify what leaves the stage.