Capabilities
AI & Automation
LLM apps, agents, and browser automation — multi-model pipelines that pull structure out of documents and drive real workflows.
What it is
The layer that ties the others together and does the work no one wants to do by hand. LLM apps and agents, browser automation, and pipelines that turn a pile of documents into structured data you can trust — multi-model, because no single model wins every task.
What we build
- LLM applications and agents, routed across models via OpenRouter.
- Browser automation at scale — agentic control over real browsers through the Chrome DevTools Protocol.
- Document and data extraction: multi-model pipelines with a rule engine that pulls structure out of unstructured files.
- Transcription and analysis pipelines on Whisper and GPT.
Stack & tools
- OpenRouter
- CDP
- Whisper
- GPT
- Gemini
- Python
Related work
AI & Automation
A browser farm and a multi-LLM rule engine for document due diligence
a due-diligence firm
A due-diligence firm needed to pull verifiable facts out of source documents that only exist behind logged-in web portals. DigitalBorders built the whole pipeline — a farm of real Chrome browsers, driven over the DevTools Protocol, that collect the documents, and a multi-LLM rule engine that extracts each fact and ties it back to where it came from.
- Neko
- Puppeteer
- Chrome DevTools Protocol
- OpenRouter
- Convex
AI & Automation
A browser kiosk that reads a gesture and a word, then brews a real espresso
a global coffee-machine maker
A global coffee-machine maker, launching a new espresso machine, wanted a stand experience where a visitor brews a real cup with nothing but a gesture and a word. DigitalBorders built it as a single browser page on an ordinary tablet — two MediaPipe vision models read the visitor's pose and the Italian "perfetto" hand gesture while speech recognition listens for the spoken word, a staged coaching loop nudges near-misses along, and on success the page reaches the machine directly over Web Bluetooth and writes the manufacturer's brew command, so the new machine pours in front of the visitor, triggered by their gesture. No native app, no attendant — it attracts, runs, dispenses, and resets itself all day.
- TypeScript
- MediaPipe Tasks Vision
- Web Bluetooth
- Web Speech API
- Whisper (WASM)