<Phi2AI/>

About Phi2AI

Logic and philosophy at the core of AI development.

Lab

// An experimental lab for AI development and analysis.

Development

Our development is grounded in diverse logical theories and philosophical inquiry. We aim to solve the challenges of modern Large Language Models (LLMs), where thinking processes are often overly complex and suffer from high token demand, by creating more streamlined and transparent architectures.

The Potential of SLMs

Small Language Models (SLMs) show immense potential for Agentic AI, as well as local and mobile device integration. We focus on one of the industry's primary challenges: enhancing the reasoning and analytical depth of these models without compromising their efficiency.

Human-AI Interaction

As human-AI interaction rapidly expands, philosophical analysis provides a vital perspective. We use these frameworks to identify critical intersections between human and machine, offering fresh insights into how this relationship will evolve.

CH-AI Implementation

Cultural Heritage Sector is one the key sectors in Digital Transformation process in decades. Every LLM development consists of some amount of digitized CH material. Also CH sector, as an experimentation playground improve and build a standards, digital access and development.

Phi2AI Logo

LLM / SLM

  • Algorithms
  • Datasets
  • Training / Fine-tuning
  • Deployment

Infrastructure

  • Local
  • Public
  • Engineering
  • Open Source

Resources

  • Cultural Heritage
  • Publicly Available
  • Synthetic Data
  • Multimodal (Text, Image, Sound)
Andrija Sagić speaking at MLC2025

Founder & Lead Researcher

Andrija Sagić

Graduated Philosophy on contemporary epistemology, Andrija has spent over a decade at the intersection of cultural heritage and advanced technology. As the Head of Digital Development Department at the "Milutin Bojić" Library and Chair of the EuropeanaTech Steering Group, his work focuses on the practical application of AI in complex data environments and philosophical insights into technology’s role in society. Certified HuggingFace AudioML and AI Agents.

With 10+ years of experience in CH sector as System Administration, FLOSS and infrastructure development, AI implementation, he leads the core experiments at Phi2AI. Founded to resolve the inefficiencies of modern LLMs, particularly for Low Resource Languages, his research currently explores the integration of Philosophy and CH to create more logical, agentic, and transparent AI systems.

Open to Collaboration

Phi2AI is an independent research initiative. Researchers, ML engineers, and institutions are welcome to reach out.

Get in →