Projects
Each one covers a different part of the stack, warehousing, orchestration, streaming, geospatial, and applied AI. Every card links through to the problem, the architecture and the code.
Personal / Portfolio project
Problem
APRA publishes superannuation performance data as unversioned Excel filings, with no API and no easy way to compare funds over time.
Key result
Weekly pipeline runs unattended end to end, gated by dbt tests and CI, feeding a live 3-page Power BI dashboard.
Personal / Portfolio project
Problem
Backtesting a forecasting model on financial time series is easy to get subtly wrong: without point-in-time discipline, the model ends up training on information it wouldn't have had yet.
Key result
A clustering key fix cut partitions scanned on date filters by roughly an order of magnitude, and Time Travel snapshots kept lookahead bias out of every backtest.
Personal project
Problem
Producing a newsletter and a companion video from breaking news end to end usually means a person doing research, writing, narration and publishing by hand, every single cycle.
Key result
The workflow runs from trigger to published newsletter and video with no manual step, and recovers from failures instead of stopping cold.
Individual project, Monash University FIT5202 Big Data Processing
Problem
A delivery-time prediction only matters if it's ready before the order needs a driver assigned, which rules out batch scoring after the fact.
Key result
Inference runs inside the stream itself, scoring available drivers per order and assigning the fastest, with live choropleth maps refreshing every 15 seconds.
Team project, Monash University
Problem
Flood, bushfire and heat risk data for a given Melbourne suburb is scattered across shapefiles, historical station records and separate forecast feeds, with no single place that turns it into a suburb-level answer.
Key result
A weighted spatial risk score (60% flood zone intersection, 40% historical fire frequency) turns five raw data sources into a single Low-to-Very-High rating per suburb.
Hackathon project, Grand Challenge
Problem
Online harm rarely shows up as a single flagged word, it spans grooming progression, deepfakes, coordinated attacks and video content that text-only moderation tools don't see.
Key result
Around 5,000 lines shipped inside a hackathon window, with 90+ detection patterns across 8 threat categories and multi-modal analysis of text, images and video.