🌟 Editor's Note
Welcome to another exciting week in the Vision AI ecosystem! We've got a packed newsletter full of insights, events, and inspiring stories from the heart of innovation.

Top Stories :

  • Prime Intellect, valued at $1 billion, has acquired Moondream, bringing founders Jay Allen and Vik Korrapati aboard to lead inference and multimodal AI development. The deal adds Photon, Moondream’s GPU inference engine, and Moondream 3.1, an efficient open vision model. Moondream Cloud is scheduled to shut down November 13. [link]

  • Stockholm-based Sortrace has raised approximately $800,000 in funding led by Almi Invest, with participation from Antler, Gorilla Capital and angel investors. Its camera-based computer vision technology monitors waste collection trucks, detecting incorrectly sorted household waste at bin and route level. The funding will support expansion beyond collection vehicles. [link]

  • Protex AI has partnered with environmental, health and safety consultancy ALL4 to help industrial organizations turn workplace video intelligence into measurable safety improvements. Protex connects existing cameras and operational systems to identify risks, while ALL4 helps implement corrective actions. The partnership initially targets manufacturing, food and beverage, forest products and waste management. [link]

Market Insider :

  • Pensa Systems has launched Actions AI, a retail execution agent that uses computer vision to inspect store shelves, identify merchandising issues and guide frontline workers through prioritized tasks. It helps reduce manual barcode scanning and auditing. Early results indicate up to 70% automation of manual work, 20% fewer stockouts and 90% fewer shelf voids. [link]

  • Australian startup Indrus has raised AU$750,000 (approximately US$526,000) in pre-seed funding led by Purpose Ventures. Its computer vision platform uses existing site cameras to track trucks, trailers and equipment across logistics yards in real time. Founded by Ryan Liebregts and Dre Siwela, Indrus plans to expand its team and deployments. [link]

  • Brazilian company ALTAVE is expanding its AI-powered safety deployments across six countries through an offshore drilling partnership. Its DROPS system detects missing tools and equipment risks, while MHER measures worker exposure in hazardous areas. Interlock can automatically slow machinery. ALTAVE reports deployments across 11 countries, including 90+ vessels and 28 rigs. [link]

  • Fainders.AI has been selected for South Korea’s 2026 Deep-Tech TIPS programme, securing up to ₩1.5 billion in R&D funding over three years. The company will enhance its Vision Check-Out (VCO) technology using multi-angle 2D images and 3D shape data to improve product recognition accuracy, particularly for overlapping and irregularly shaped items, while reducing hardware costs. [link]

  • Eyedaptic launched Eva, an agentic AI visual assistant integrated with its EYE7 Vision Enhancing Smart Glasses. Combining computer vision and natural-language understanding, Eva interprets user intent and automatically adjusts visual settings. Company testing reported over 99% reading accuracy, 97% command accuracy, and support for up to 99 languages, helping people with visual impairments. [link]

  • SentriOS has joined Axis Communications’ Technology Integration Partner programme, bringing AI-powered security capabilities to existing Axis cameras. Its vision-language models understand entire scenes across construction, utilities and public safety environments. The company reports over 95% fewer false alarms while maintaining over 99% recall, alongside automated audio warnings and verified emergency dispatch. [link]

Pensa Actions AI

Research Spotlight :

  • Ramblr.ai’s BARISTA dataset contributed to research behind StreamVibe, a proactive video-agent framework designed to recognize explicit, implicit and changing user intent. StreamVibe introduces a benchmark with 4,100 evaluation samples and streaming model that decides when to assist, improving response timing and answer quality by 10.7%; hierarchical memory adds 12.3%. [link]

  • Researchers developed SFMDet, a YOLOv12-based computer vision architecture for detecting objects across complex construction sites. Combining state-space scanning, multi-scale cross-attention and task-specific prediction modules, it achieved 72.6% mAP, outperforming its baseline by 4.3 percentage points. Tested on 5,000 images, it processed up to 287 frames per second on an RTX 4090. [link]

Community Spotlight:

  • In the latest Bloomberg Tech Disrupters’ podcast, Avathon CEO Pervinder Johar discussed how Physical AI is transforming industrial operations through autonomous systems. By converting enterprise data into actionable knowledge, Avathon aims to improve operational performance and extend the lifespan of critical infrastructure across industries.

  • Focal won Emerging Retail Technology Innovation of the Year at the 2026 RTIH Innovation Awards. Its computer vision platform helps retailers identify hidden inventory, reduce shelf-audit labor, improve product availability, and automate replenishment workflows.

  • Pony.ai and GAC Lingcheng have begun mass production of their T9 Robotruck, featuring L4 autonomous driving, 360° perception through 22 sensors, 70% lower ADK BOM costs, and 10% reduced energy consumption.

Till next time,

Vision AI weekly