AI & ML interests

None defined yet.

Recent Activity

Vicens  updated a Space about 15 hours ago
Scalai/README
Vicens  published a Space about 15 hours ago
Scalai/README
View all activity

Organization Card

SCALAI: Advanced Artificial Intelligence & Algorithmic Solutions

Welcome to the official Hugging Face organization of SCALAI.

We are an expert Artificial Intelligence software company based in Spain, dedicated to designing and implementing advanced algorithmic solutions for complex and unique projects. Our mission is to bridge the gap between cutting-edge AI research and robust deployment in the business and tech world.

Who We Are

At SCALAI, we believe that state-of-the-art AI should not be limited by prohibitive hardware barriers. Although we are recognized for our work in optimizing Large Language Models (LLMs), our core activity encompasses the development of comprehensive solutions in Computer Vision, Machine Learning, and Predictive Analytics. We transform heavy, resource-intensive models into surgically precise and highly efficient systems, delivering transformative results for industries such as Industry 4.0, Healthcare, Finance, and Retail.

Empowering the Entrepreneurial Ecosystem and Deep Tech Startups

One of our highest-impact strategic activities is our close collaboration with the innovation and entrepreneurial ecosystem. At SCALAI, we help Deep Tech companies build solid commercial products based on their core technologies.

We actively facilitate the scaling phase of disruptive startups, guiding founders and technical teams from Proof of Concept (PoC) to full production deployment. By optimizing their algorithms and AI software architectures, we ensure that startups can grow in a scalable, robust, and efficient manner—preparing them to compete at the highest technological level and attract investment.

Our Open Source Philosophy

We are deeply committed to the democratization of AI. We research and develop novel methodologies for model compression, sparsity, and domain specialization, aiming to make massive models accessible on standard enterprise hardware (such as the NVIDIA L40S). We use this space on Hugging Face to share our most efficient models and architectural breakthroughs with the open-source community.

Featured Models

  • ScaLite-60B-Coder: A highly specialized, bilingual (English/Spanish) Mixture-of-Experts (MoE) model optimized for strict logical reasoning, Python programming, and mathematics. Utilizing our novel Activation-Guided Sparsity and Cross-Domain Router Healing methodologies, we distilled a dense 120B MoE into a 60B active parameter footprint that runs natively on a single 48GB GPU without sacrificing its reasoning capabilities.

Our Areas of Expertise

  • Startup Scaling & Deep Tech: Technical guidance, AI software architecture, and algorithmic optimization to help emerging businesses grow.
  • Large Language Models (LLMs): MoE architectural optimization, activation-guided pruning, router healing, and advanced quantization (MXFP4).
  • Computer Vision: High-precision object detection, segmentation, and automated visual inspection systems.
  • Predictive Analytics: Advanced algorithmic modeling for complex datasets and decision-making.
  • Custom AI Software Engineering: End-to-end deployment of Artificial Intelligence solutions into production environments.

Connect with Us

Whether you are looking for a specialized AI partner to scale your startup, develop a unique Computer Vision project, or simply want to discuss our open-source models, we would love to connect with you.


Unlocking Goals!.

datasets 0

None public yet