Transfusion of transformers and state space models

Boson AI specializes in AI-driven simulation and optimization, focusing on multi-modal learning, generative modeling, and structured state-space architectures. The company develops AI models that integrate text, audio, and structured signals to enable high-fidelity simulation and synthesis across domains like speech processing, acoustics, and human-AI interaction. Through this project, Boson AI aims to address key challenges in multi-modal fusion and instruction-driven audio synthesis. The intern will help bridge the gap between discrete and continuous modalities by developing a unified learning framework. The intern will also work on capturing long-range dependencies in audio signals to improve synthesis coherence and quality. Additionally, the intern will enhance user control over AI-generated audio, enabling customization, interpolation, and real-time adaptation, while also ensuring scalability for diverse linguistic, tonal, and contextual inputs. This project will strengthen Boson AI’s AI capabilities in multi-modal learning, opening new business opportunities in media, virtual assistants, entertainment, and accessibility technologies. It will also contribute to intellectual property advancements, positioning Boson AI as a leader in generative audio synthesis. By supporting commercialization efforts such as licensing models and API development, this work will align with Boson AI’s mission to create scalable and adaptive AI technologies with long-term industry impact.

Faculty Supervisor:

Vardan Papyan

Student:

Partner:

Boson AI

Discipline:

Computer science

Sector:

Education; Information and cultural industries

University:

University of Toronto

Program: