A Portfolio · 2026

Hi, I'm Steve.

This is where I keep the things I'm building with AI right now. Some of it is BizOps stuff from my day job. A lot of it is just side projects where I'm trying to learn by actually shipping things. Some are live, some are still pretty rough. But it's an honest look at where I'm at with all of this.

📍 New York Resume PDF →
Stephen Fung
Stephen Fung

Selected Work

7 PROJECTS

Two threads here. The work side is what I've shipped at Shippo. The side projects are how I learn by building stuff I'd actually use myself. Click any tile for the full story.

From work

BizOps and GTM

From the side

How I learn by building
A personal system · how I work

My AI Operating System

A simple, inspectable system that connects every AI tool I use, so my work has memory, my decisions get written down, and nothing has to start cold.

One markdown vault is the source of truth. Claude Code, Codex, and a nightly sync agent all read and write to the same place.

Obsidian + Dataview In production daily
read the story →
An AI advisor · for travelers

AwardReserve

An AI assistant that tells you the smartest way to spend your credit-card points (flights, hotels, transfer partners) without making things up.

I built it around a hand-curated knowledge base after the AI confidently invented a hotel category. Now it's all provenance-checked data with real verification dates and grounded answers.

Claude API Pre-beta · live preview
read the story →

Try the live preview at pointsiq2.netlify.app ↗ (temporary URL)

A planning tool · for everyone

FIRE Calculator

A friendly retirement-planning website. Punch in your numbers and it tells you whether you're on track. An AI assistant walks through what to do if you're not.

Under the hood: 5,000-path Monte Carlo simulation, Roth ladder modeling, withdrawal guardrails, and a Claude Haiku advisor that reads every input live.

Claude Haiku Stripe-monetized
read the story →
An automation · for job hunters

LinkedIn Job Pipeline

A quiet helper that scores LinkedIn jobs against my goals, writes a tailored resume and cover letter for each one, and drops everything into a Drive folder for me to review.

Runs entirely on a free local AI model. No API keys, no per-job cost. A simple dashboard handles approve / skip.

Local AI · Ollama $0 to run
read the story →
A homelab · for the household

Self-Hosted Media + Photos

A private streaming service and a photo backup that lives on my own hardware. Family asks for a movie, it shows up in Plex an hour later, no streaming-service hopping required.

Seven Docker services on a UGREEN NAS, glued together with Tailscale so I can reach all of it from anywhere. No public internet exposure.

Docker + Tailscale Live
read the story →

A bit about me

THE LONGER VERSION

I'm a BizOps person. Have been for about 15 years across various B2B SaaS and fintech companies. But honestly, the part of the job I like most is the shipping part. Scoping something on Monday, wiring it up by Friday, and writing whatever needs to go around it after.

The projects on this site are mostly me trying to figure out where AI actually fits into all of that. Some of it is at work, where I'm trying to use AI without breaking anything. Some of it is on the side, where I just build things I'd want to use myself. Either way, the goal is the same: keep things simple, write down what matters, and don't add cleverness for its own sake.

Outside of work I tinker. I've also accidentally turned my childhood Pokemon card collection into something a little too overwhelming, and I'm using it as an excuse to learn how ecommerce actually works.

How I work

  • I document important decisions where people need to see them first, then store everything as markdown so my LLMs can pull from it later
  • I prefer durable artifacts over vague advice
  • I automate things only after I've done them by hand at least once
  • I'd rather reduce complexity than add cleverness
  • Local-first when it's practical, cloud-first when it's smarter

What I reach for

  • AI: Claude (Sonnet, Haiku), Anthropic SDK, OpenRouter, Ollama
  • Infra: Docker, Tailscale, NAS hardware, macOS LaunchAgents
  • Tools: Claude Code, Codex CLI, Cursor, Obsidian + Dataview

Still learning (as I go deeper into the LLM space):

  • Web: Vanilla HTML/CSS/JS, Next.js, Netlify Functions
  • Data: Markdown + YAML, JSON caches, Supabase, SQLite, Databricks SQL

Get in touch

Email is the best way to reach me. Happy to walk through any of the projects above, or just say hi.