🌟 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:
MountAIn partnered with Alif Semiconductor to deliver cloud-grade computer vision on ultra-low-power Ensemble and Balletto processors. Their “AI Booster” middleware enables one-click Python-to-silicon deployment, cuts AI model memory usage by 3x, and reduces development cycles from 12 months to minutes for battery-powered edge AI devices under 100mW. (link)
BHP deployed an AI-powered vision system across its iron ore operations to predict equipment failures, reduce unplanned downtime, and improve production efficiency. The computer vision platform analyzes conveyor belts and mining infrastructure in real time, boosting ore output, lowering maintenance costs, and accelerating automation across large-scale mining operations. (link)
Pirelli partnered with Univrses to integrate AI-powered 3DAI computer vision into its Cyber Tyre platform. The system combines tyre sensor data with onboard camera vision for real-time road monitoring, ADAS, and autonomous driving applications. Pirelli also acquired a 30% stake in Univrses to accelerate software-defined vehicle innovation. (link)
Swiss startup Moonlight AI raised $3.3M in Seed funding to scale its AI-powered cancer diagnostics platform. The company uses computer vision to extract genomic biomarkers directly from routine blood and cytology slide images, enabling faster, lower-cost diagnostics for hematology and oncology applications. Investors include Lotus One Investment, VP Venture Partners, and MEDIN Fund. (link)
Startup Spotlight
Moonlight AI is a Swiss healthtech startup building AI-powered cancer diagnostics using computer vision.
Founded in 2022 by Christian Ruiz, Nicole H. Romano, and Stefan Habringer, the company analyzes routine blood, bone marrow, and cytology smear images to detect genomic biomarkers without relying heavily on expensive Next-Generation Sequencing (NGS).
Its platform combines whole-slide imaging with multimodal genomic datasets to help hematologists and pathologists diagnose diseases such as myelodysplastic syndrome (MDS), chronic lymphocytic leukemia (CLL), and non-small cell lung cancer faster and at lower cost.
The software integrates directly with existing lab imaging workflows, enabling faster turnaround times and scalable precision oncology diagnostics.
In 2026, the startup raised $3.3M in seed funding to expand its proprietary dataset, grow its clinical consortium, accelerate regulatory approvals, and scale commercialization globally.
Paper to Factory
Prefill-Time Intervention for Mitigating Hallucination in Large Vision-Language Models
A new paper, Instant Colorization of Gaussian Splats, introduces a visibility-weighted least squares framework that projects 2D image information directly back onto 3D Gaussian splats — without costly gradient descent optimization.
The key idea: solve Gaussian colorization analytically using the normal equation, while explicitly handling:
• occlusions
• transparency-aware alpha blending
• view-dependent spherical harmonics
The system enables:
relighting 3D scenes from new illumination captures
projecting DINOv2 features into 3D semantic maps
SAM2-powered 3D segmentation without additional training
Results show up to 10× faster convergence than Adam/AdamW baselines while achieving better reconstruction metrics.
🏆 Community Spotlight:
Till next time,
