DEV Community

Cover image for Building Neo v0.1: Desktop Screen Perception & Automation with LuxurAI Cowork
Achyut Srivastava
Achyut Srivastava

Posted on

Building Neo v0.1: Desktop Screen Perception & Automation with LuxurAI Cowork

What We Built Today

Today, we shipped Neo v0.1, our in-house vision-language model, designed for real-time desktop screen perception and automation. Neo v0.1 runs on our dedicated dual-GPU cluster, processing live desktop frames at 15 fps, detecting UI elements, and performing OCR for text extraction.

The Problem We Solved

Desktop automation requires precise visual grounding and low-latency processing. Most solutions rely on expensive APIs or lack native support for Indian languages and high-DPI setups. Neo v0.1 addresses these challenges with a lightweight, localized vision engine.

How The Architecture Works

Neo v0.1 combines:

  • Screen Capture Pipeline: Continuously captures desktop frames and downsamples them for processing.
  • UI Element Detection: Uses bounding boxes to identify active application states and UI components.
  • OCR & Text Extraction: Extracts text from live frames for task execution.
  • Human-in-the-Loop Safety Gate: Ensures every high-risk action requires explicit user approval.

Transparent Economics

LuxurAI Cowork is priced at ₹0.25 per credit (3 LC/minute ≈ ₹0.75/minute), with no subscriptions or hidden fees.

How We Can Make This Better

We’re actively optimizing Neo v0.1 for multi-monitor setups and high-DPI environments. Feedback from the AI community is invaluable.

Try LuxurAI Cowork

Test the beta and share your feedback: https://luxurai.in

What do you think? How can we improve Neo v0.1? Let’s build something powerful together!

Top comments (0)