Things I've Built!

Co-founding TerraMind — turning raw atmospheric and field data into calibrated weather and agronomic probabilities, then into block-level, dollar-denominated decisions. Built a gradient-boosted meta-model that blends multiple forecast providers (ECMWF, GFS, ICON, MetService) by learned per-provider skill, feeding an XGBoost bias-correction layer and a 1000-sample Monte Carlo engine that produces calibrated, empirically-validated probability distributions for frost, disease, yield, and irrigation risk across 13 NZ regions. Layered an AI reasoning system on top — powered by Gemma 4 — that plans the season, triages what threatens it, and compounds seasonal intelligence per block, with every LLM output backed by a deterministic fallback so the system stays correct even when the model isn't. Unified ingestion pipeline across Open-Meteo, MetService, LINZ, and Sentinel-2 satellite data, feeding a PostGIS observation schema.

Built a full-stack SaaS platform that helps freelancers and agencies protect against scope creep. Developed AI-powered request analysis using LLMs to evaluate client requests against project scope, automatically estimate costs and effort, and generate client approval workflows. Integrated with Slack for request capture and GitHub for issue creation and tracking. Built with Next.js App Router, TypeScript, React 19, Supabase (Auth, Postgres, RLS), OpenRouter AI, Slack OAuth, and GitHub App API. Created at SaaSathon 2026, placed 3rd.

Built a full-stack quantitative research platform for backtesting and comparing systematic equity/ETF strategies against benchmarks. Developed six built-in strategies, including machine learning models using Ridge regression and LightGBM with walk-forward validation, plus downloadable tearsheets, portfolio analytics, and a side-by-side comparison dashboard.

Built an AI-powered, community-driven platform for reporting campus hazards, enabling faster delivery of critical information to security staff and centralizing safety data to support university decision-making. Awarded 1st place at the Deloitte × ServiceNow × University of Canterbury Hackathon during Tech Week NZ 2025.

Developed a responsive portfolio website using Next.js, TypeScript, and Tailwind CSS, implementing Framer Motion animations, dark/light mode functionality, and performance and SEO optimizations.