Skip to content
Back to work

Platform

Kairo

Role
Full stack engineer at Grow_in
Period
Mar 2026 to Jul 2026
Stack
  • React
  • TypeScript
  • Tauri
  • FastAPI
  • OmniRoute
  • edge_tts
  • Docker
  • CI/CD

Kairo is an AI video generation platform that ships as a desktop app. React and TypeScript inside Tauri, talking to a FastAPI backend.

The pipeline runs in three stages. A Writer produces a ten scene script through OmniRoute LLM routing. A TTS module narrates it and keeps the word boundary timestamps edge_tts returns, so the narration is not just audio but audio with known positions in it. An Aligner then reads every scene against a 2.25 words per second baseline and flags the ones that will not sit in their slot.

Writer10 scenes, OmniRouteTTSedge_tts word timingsAligner2.25 words / secEDLfrom JSON scenesone click, raw footage in, film out

Drag the diagram sideways to see the rest

Writer, TTS, Aligner, EDL. The word timings edge_tts hands back are what let the Aligner check a scene against 2.25 words per second before anything is cut.

The last stage writes the edit. Scene definitions go in as JSON and an EDL comes out, so the cut is assembled from data rather than dragged together by hand. Manual production effort on a video dropped by 80%.

In the same run I built Fable, a Shopify app with three custom components, and dockerised the stack behind CI/CD.