<aside>
Rethinking development through an AI-first lens.
Creative Embedded & AI-Native Prototyping
I make AI that runs on your own device instead of the cloud, and small gadgets that respond to the world through sensors. I came to this from psychology rather than engineering, and I learn by building the things I'm curious about — lately across small microcontrollers (ESP32) and single-board computers (Raspberry Pi), figuring it out as I go.
</aside>
<aside>
Primary Tools: Antigravity IDE · Claude Code · ChatGPT · VS Code · Arduino IDE · Ollama
</aside>
<aside> <img src="/icons/slideshow-play_gray.svg" alt="/icons/slideshow-play_gray.svg" width="40px" />
</aside>
<aside> <img src="/icons/movie-clapboard_gray.svg" alt="/icons/movie-clapboard_gray.svg" width="40px" />
https://youtube.com/shorts/aOwOBazgZFE?feature=share
<aside> <img src="/icons/compose_gray.svg" alt="/icons/compose_gray.svg" width="40px" />
My first real app, built by directing AI through the whole process — design, code, and launch with ChatGPT (I did not know how to use AI IDE yet). I wanted to find out what happens when you treat AI as a team to orchestrate rather than an occasional helper. This is where "AI-native" building first started to click for me. Read the detailed showcase →
Dream Home Vibe: My First Experiment as an AI-Native Orchestrator
</aside>
</aside>
<aside> <img src="/icons/movie-clapboard_gray.svg" alt="/icons/movie-clapboard_gray.svg" width="40px" />
https://youtube.com/shorts/u6rP2D4xVG8?feature=share
<aside> <img src="/icons/compose_gray.svg" alt="/icons/compose_gray.svg" width="40px" />
A voice assistant that works completely offline — you speak, it responds, and nothing leaves your machine. Most voice assistants send what you say to a company's servers, and I was curious whether a fully private one could still feel quick and natural. So I tried building it.
Read the detailed showcase →
Kai Lite: Building a Privacy-First Voice Assistant Through AI Orchestration
</aside>
</aside>
<aside> <img src="/icons/movie-clapboard_gray.svg" alt="/icons/movie-clapboard_gray.svg" width="40px" />
<aside> <img src="/icons/compose_gray.svg" alt="/icons/compose_gray.svg" width="40px" />
Most AI assistants reset the moment you close them. Kai Desktop is my attempt at the opposite—it keeps a long-term memory and reflects on past conversations to adjust how it responds. I also built mechanisms to protect my privacy: the core agent runs entirely on a private home desktop so your personal data never touches the cloud, and I can swap out the local LLM whenever I want. When I need more power, I can intentionally leverage cloud AI via API commands. It's the idea I keep coming back to: an AI that grows alongside you instead of starting over each time. 📄 Read the showcase →
</aside>
</aside>
<aside> <img src="/icons/movie-clapboard_gray.svg" alt="/icons/movie-clapboard_gray.svg" width="40px" />
<aside> <img src="/icons/compose_gray.svg" alt="/icons/compose_gray.svg" width="40px" />
Groundmarker began with a simple problem: an inspector should not have to remember which camera-roll photo belongs to which spot on a hand-drawn sketch. During an inspection, the app lets them draw the site, mark a finding, and take photos directly from that marker—automatically preserving the connection between what was found and where. Once the inspection is complete, Groundmarker turns everything into an organized, client-ready PDF. Built first for crawlspace inspections, it can support any field where visual evidence and precise location need to stay connected from the jobsite to the final report.
📄 Read the showcase →
</aside>
</aside>
<aside> <img src="/icons/light-bulb_gray.svg" alt="/icons/light-bulb_gray.svg" width="40px" />
Dotchi is a small desk creature I'm building on an ESP32 — a cheap, tiny computer-on-a-chip behind a lot of DIY gadgets — with a round, glowing face. He notices you a little like a pet does: a touch on the screen, a sound in the room, a tilt when he's picked up, and, the part I'm most interested in, whether you're actually nearby — sensed through a small radar rather than a camera. Dotchi's only job is to feel like a creature. I recently replaced an early shortcut — one touch used to make him assume you'd stay for 90 minutes — with radar that actually reads whether you're still at the desk. I am working on the Machine Learning part to build him an offline presence currently.
An idea I'm still circling: a small tabletop garden where a mythical creature appears to float in mid-air, using an old glass-reflection trick to give it a soft, almost-holographic depth. It would run on a single-board computer (Raspberry Pi), and the whole point is wonder — making something that feels a little magical and alive sitting on a shelf. Early days, but it's the kind of thing I can't stop thinking about.
</aside>
For any inquiry, please email emilylindev [at] gmail [dot] com