Full-Stack Developer

Building web applications with a focus on serverless architecture, real-time data pipelines, and developer tooling.

• Projects
  • bomweather.com
  • aitrackerstats.com
  • obsidian-charts
  • mytart
  • whoistld
• Tech Stack Summary

Real-time Australian weather observations, forecasts, and radar imagery sourced from the Bureau of Meteorology.

• Framework: SvelteKit (Svelte 5) + TypeScript
• Styling: Tailwind CSS 4
• Database: Neon (serverless PostgreSQL) via Prisma + Kysely
• Cache: Upstash Redis
• Maps: Leaflet interactive radar
• AI: Vercel AI SDK for weather insights
• Analytics: mytart
• Deployment: Vercel Edge Runtime (syd1 region)

Architecture: Cron jobs fetch BOM observation, warning, and forecast data, cache it in Redis, and serve it through API endpoints. The frontend renders a region-switchable dashboard with live observations, 7-day forecasts, 3-hour forecasts, rain probability, and radar. Region selection persists via localStorage, defaulting to QLD/Brisbane.

API surface:

• /api/observations/{region}/{city} — current weather observations
• /api/warnings/{region} — active weather warnings
• /api/forecasts/{region}/daily — 7-day forecast
• /api/forecasts/{region}/threehour — 3-hour hourly forecast
• /api/forecasts/{region}/rain — rain probability forecast

AI-powered web analytics platform providing traffic analysis, insights, and team collaboration.

• Framework: SvelteKit 2 (Svelte 5) + TypeScript
• Styling: TailwindCSS
• Database: Neon PostgreSQL via Prisma + Kysely
• Cache: Upstash Redis
• Auth: Lucia Auth with OAuth (Google, GitHub)
• Email: Resend
• Error Tracking: Sentry
• Visualizations: Chart.js
• Analytics: mytart
• Deployment: Vercel

Features: analytics dashboard with charts, historical data archives, custom domain verification, multi-user team management, subscription billing, AI-powered traffic recommendations, and privacy/compliance tooling.

Obsidian plugin for creating interactive charts directly in your notes using Chart.js. Originally a fork, extensively refactored and enhanced with new features, a complete test suite, and modernized architecture.

• Language: TypeScript (strict)
• Plugin Framework: Obsidian API + Svelte UI components
• Charts: Chart.js v4 with Sankey, Financial (Candlestick/OHLC), and 60+ Chart.js options
• Testing: Jest with comprehensive mocks (Obsidian, Chart.js, DOM APIs)
• Build: Rollup, automated releases via GitHub Actions
• Package: obsidian-charts

Key Features:

• YAML-based chart definitions in ```chart code blocks with full type safety
• 8 Chart types: Bar, Line, Pie, Doughnut, Radar, Polar Area, Sankey, Candlestick, OHLC
• Table linking: Bind charts to Markdown tables via block IDs — data stays in sync
• Date auto-transpose: Tables with date columns automatically format as time series
• Best fit lines: Linear regression computed against data series via bestFit modifier
• Graphical creator: Command palette UI for building charts without typing YAML
• Chart → Image: Export rendered charts as PNG/JPEG with customizable quality
• Dataview integration: window.renderChart() API for programmatic rendering in DataviewJS

Architecture & Refinements (vs original):

• Completely rewritten chart rendering pipeline (chartRenderer.ts) with proper dataset preparation, color management, and Chart.js option handling
• Table-to-chart conversion (chartFromTable.ts) with empty-cell handling, layout detection (rows/columns), and cross-file linking support
• Modernized plugin lifecycle: Proper chart destruction, resize handling, and metadata change detection with debouncing
• YAML preprocessing utility to handle Obsidian's YAML parsing edge cases and type coercion (strings → numbers, empty → null)
• Robust test suite (700+ tests) with Jest mocks for Obsidian, Chart.js, Sankey, Financial, date adapters, and color libraries
• Release automation: Version validation, semver enforcement, and conditional tag/draft/prerelease handling in GitHub Actions
• UI improvements: Collapsible settings sections, helper modals, custom picker integration, and dark mode enhancements
• Code quality: Type-safe interfaces (types.d.ts), separated concerns (rendering vs table logic), and comprehensive error handling

Modifiers & Options: 40+ chart modifiers including stacked, fill, tension, stepped, spanGaps, beginAtZero, axis controls (xMin/yMax), per-series best fit, and full axis title/display configuration.

Multi-Yield Tracking & Analytics Relay Tool — a framework-agnostic analytics library that fans out events to 17 providers simultaneously via direct HTTP calls.

• Language: TypeScript (strict)
• Build: tsup (dual ESM/CJS output + DTS)
• HTTP: axios with retry interceptor + exponential backoff
• Fingerprinting: ThumbmarkJS (lazy-loaded, browser-only)
• Parsing: ua-parser-js (bot detection)
• Package: npm install mytart

Providers: Google Analytics 4, Google Ads, Mixpanel, Segment, Amplitude, Plausible, PostHog, Meta Pixel, Microsoft Clarity, Hotjar, Heap, TikTok, Snapchat, Twitter/X, Reddit, Pinterest, Microsoft Ads — each supporting browser and/or server modes.

Key architecture decisions:

• No SDK wrappers: direct HTTP via axios across all server-mode providers
• Parallel dispatch: all providers called concurrently via Promise.all
• Per-provider enabled flag: providers off by default
• Provider groups: route events to marketing, product, infrastructure, or physical subsets
• Standardized event taxonomy: 36 standard event names auto-mapped to each provider's native conventions (36+ mappings per provider)
• Cross-provider linking: auto-captures click IDs (gclid, fbclid, ttclid, twclid, msclkid, li_fat_id) and cookies (_fbp, _fbc, _ga) for unified attribution
• Browser fingerprint: ThumbmarkJS generates a stable cookieless device identifier as anonymousId
• Retry + DLQ: configurable exponential backoff with in-memory dead-letter queue and replay
• Global consent: single consent: true/false flag configures GA Consent Mode v2 + Meta Pixel grant/revoke
• Never throws: errors returned as TrackResult with success: false

TypeScript package for parsing domain names and extracting TLD characteristics — WHOIS privacy support, registration restrictions, DNSSEC, IDN, and category classification. All data embedded locally, zero network requests.

• Language: TypeScript (strict)
• Validation: Zod schemas for all input/output types
• Build: tsc (CJS output + DTS)
• Dependencies: zod, axios
• Package: npm install whoistld

API surface:

• parseDomain(input) — parse a domain into TLD, domain, subdomain, and TLD characteristics
• supportsWhoisPrivacy(tld) — check if a TLD supports WHOIS privacy
• getTldData(tld) — get full characteristics for a TLD
• getAllTlds() — list all known TLDs
• isKnownTld(tld) — check if a TLD exists in the database

TLD characteristics tracked: whoisPrivacySupport, registrationRestrictions (none/governmental/educational/organizational/geographic/reserved), dnssecSupport, idn, category (generic/country-code/sponsored/infrastructure/test/generic-restricted), and description.