- Clinical AI 1: Built a world-first cry detection system for nurses to better attend to newborns in the neonatal intensive care unit in collaboration with clinical stakeholders. Single-handedly improved the AI component (a CNN trained on audio spectograms) from a TRL 4 (50% precision) to a TRL 8 (96% precision, 97% recall). Built an additional codebase to scalably ingest training data from the hospital (over 1TB) and annotate training data 5x faster.- Clinical AI 2: Developed an experimental AI pipeline that performs automated General Movement Assessment (used to assess cognitive development) of a newborn baby via video recording. Involved pose estimation models (e.g. OpenPose and Google's MediaPipe), creating an API endpoint and deployment using Docker and AWS Lambda.- Gap Selling: Built a sales funnel to sell research licenses of Neolook software, resulting in 2 paid opportunities. - Product Management: Designed/Built a ICU data solution from ideation to validation with clinicians. In addition, built performance reports and visualised user feedback. - Medical Device Regulation (MDR): Researched/Prepared regulatory filings to CE-mark Neolook's products in accordance with EU MDR software-as-a-medical device rules and relevant ISO standards.

- Big Data Wrangling: Built pipelines to process netCDF model output from 34 global & regional climate model simulations for analysis by world-leading climate scientists. The dataset is in the process of being published. - Climate Research: Wrote a research paper on the statistical influence of regional down-scaling schemes on how global climate models resolve extreme precipitation over time (supervisor Prof. Lisa Alexander). - Stats \& Data Visualisation: Programmed complex climate algorithms for use in UNSW course CLIM3001 tutorials.

- Computer Vision Development: Built an digital artwork recommender system using OpenAi's CLIP model for image encoding and KNN for inferencing a vector database of encoded images stored on Relevance AI. - Cloud Deployment: Hosted the entire pipeline on AWS using S3, Lambda and API Gateway.

- Risk Strategy: Designed a limit increase initiative projected to make $Xm+ in additional revenue for AMEX.- Big Data & Analytics: Automated interactive monthly risk reports saving 2 days of manual work, built a widely used Excel bank statement scraper to easily monitor accounts and decisioned 20+ credit limit escalations.- Thought Leadership: Convinced my risk team to incorporate a customer's expected profitability - not only their risk of default - into credit limit calculations.

In this 6-week summer scholarship I worked at the Climate Change Research Centre UNSW under Prof. Lisa Alexander and Dr Margot Bador comparing how state-of-the-art dynamical climate models scale one specific measure of extreme rainfall into the future. This was an amazing opportunity to self-learn a range of technical skills while undertaking original climate research. Achievements:- Obtained rigorous experience managing climate data (netCDF files), extracting the desired climate indices and generating geo-spatial plots (using specialised Python libraries). This was my first interaction with genuinely BIG data; the files are churned out by supercomputers that simulate the entire planet's climate every 30mins for hundreds of years!- Wrote shell scripts to automate the file management and production of plots. - Having finished the 6-weeks worth of work in 3-weeks, the scope of my work was extended to compare how the extreme rainfall index (rx1day) scaled across land/sea and seasons. - Wrote a 12-page scientific report on my research and findings. Problem Overview:"While global climate models (GCMs) remain our best tool for investigating the Earth’s system response to anthropogenic forcings, their spatial resolution (generally hundreds of km) is much coarser than the scales of the key processes leading to precipitation extremes (e.g. intense convective rainfall events). Therefore, parametrizations are necessary and the simulation of precipitation is not explicitly resolved in models. Spatial resolution is finer in regional climate models (RCMs) (generally tens of km), which is expected to improve the simulation of precipitation extremes that are very sensitive to spatial contrasts and topography.How the output of global and regional models scale with respect to the other is rarely compared. In particular, it is unclear how the future changes in precipitation extremes from large ensembles of regional climate models compare to those from global models."

- Co-founded a tutoring program designed for students with special learning needs.- Facilitated the personalised learning of 15+ students weekly.

- Used data scraping (Selenium) to build the first search engine for provider services using NDIS taxonomy.- Researched potential investors and introduced the seed round’s lead investor to Provider Choice's founders.- Tracked operational efficiency of invoice processing and implemented processes to improve processing efficiency by ~60%.- Market Analysis: Generated 24 capability documents on competitors and a set of indicators to determine their performance.- Acquired customer service skills by resolving customer complaints.