Research at Horton

We are committed to advancing the frontier of artificial intelligence through foundational research and a commitment to open science.

Core Research Areas

Large-Scale Model Pre-training

Developing novel architectures and training techniques to build more efficient and powerful foundation models with a focus on hybrid Mixture-of-Experts (MoE) systems.

Agentic AI & Tool Use

Pioneering research into autonomous agents that can reason, plan, and execute complex multi-step tasks by dynamically using tools and APIs.

AI Privacy & Security

Advancing techniques like federated learning, differential privacy, and homomorphic encryption to build inherently private AI systems and safeguard against adversarial attacks.

AI Safety & Alignment

Focusing on interpretability, controllability, and robustness to ensure our models are safe, beneficial, and aligned with human values.

Efficient On-Device Intelligence

Researching quantization, model compression, and specialized architectures to run powerful AI models directly on edge devices without compromising privacy or performance.

Multimodal Foundation Models

Building models that can natively understand and process information across different modalities, including text, images, audio, and video, to create a more holistic understanding of the world.

Our Philosophy

Our work is driven by a curiosity to understand the fundamental principles of intelligence and a desire to build AI that is safe, beneficial, and aligned with human values. We believe in a collaborative and open approach to research, actively partnering with academic institutions and the broader AI community to accelerate progress and ensure our work has a positive impact.

Publications

Our contributions to the field will be shared here. We are in the process of preparing several papers for publication and look forward to sharing our findings with the community soon.