https://steefancontractor.github.io/tags/generative-ai/Generative AIHugo Blox Builder (https://hugoblox.com)en-usWed, 29 Apr 2026 15:51:51 +1000https://steefancontractor.github.io/media/icon_hu_2f428faf7e1b7403.pnghttps://steefancontractor.github.io/tags/generative-ai/https://steefancontractor.github.io/project/countercurrent/Wed, 29 Apr 2026 15:51:51 +1000https://steefancontractor.github.io/project/countercurrent/<!-- CounterCurrent is a startup spun out of UNSW that helps shipping operators make better decisions at sea through AI-driven ocean forecasting and route optimisation. Since 2024 I have been helping build the company as ML/AI Lead, working across both frontier modelling and the broader technical platform. A major part of my work has been training large-scale ocean forecasting models on AWS infrastructure, including SageMaker AI, HyperPod Slurm clusters, and FSx Lustre. This has involved developing and testing foundation-model-style approaches for spatiotemporal forecasting using modern generative modelling methods such as flow matching and score-based diffusion. My forecasting model embodies a new AI-driven approach to data assimilation: instead of applying a traditional statistical assimilation pipeline, I nudge generative trajectories toward observations throughout the rollout process to produce data-assimilated outputs. This is aimed at learning high-resolution behaviour from physical model simulations while staying grounded in lower-resolution observational and reanalysis data. Beyond modelling, I have helped build much of the wider CounterCurrent product stack, including the web application and route-optimisation systems used for maritime decision support. I also work on technical roadmap development, supervision and mentoring, grant and program applications, and conversations with customers, partners, and investors. -->