Runhouse (Winner) and CuraVoice (Community Choice Award)
The 2025 PyTorch Startup Showcase highlighted early-stage teams building real-world AI applications and creating space for founders to connect with investors, customers, and engineering talent. Runhouse, presented by Donny Greenberg, was this year’s Showcase Winner, and CuraVoice, presented by Shrey Modi, received the Community Choice Award.
Startup Showcase Winner: Runhouse
Runhouse provides infrastructure tooling that helps AI infrastructure teams manage and grant access to their compute resources for ML engineers, researchers, and production workloads. The team also recently introduced Kubetorch, an open source Pythonic system for executing tasks on AI compute clusters with speed and fault tolerance.
Connect with Runhouse on LinkedIn, GitHub, and X
How Runhouse Uses PyTorch
Runhouse’s mission is to connect the ML stack with infrastructure like Kubernetes to improve developer experience, scaling, and fault tolerance. PyTorch is central to that integration.
PyTorch Distributed is a key component of the team’s support for distributed training frameworks. In one example, a user deploying diffusion LLM work through a Chinese RL library was able to get the system running on Kubernetes within a day because the library used PyTorch Distributed under the hood.
Why Runhouse Chose PyTorch
Runhouse relies on PyTorch across training, inference, evaluation, profiling, and data processing. The team points to the strength of the PyTorch community and ecosystem as a decisive factor in choosing it as the foundation for their interoperability work across ML and infrastructure.
What’s Next for Runhouse
Runhouse was the first to offer Monarch support on Kubernetes and is exploring ways to bring more PyTorch libraries, including torchcomms, torchforge, and torchstore, to Kubernetes. The team is also developing OpenEnv to bring RL sandboxes to Kubernetes and simplify workflows that currently require complex container setups.
Community Choice Award Winner: CuraVoice
CuraVoice builds AI-powered voicebots that help healthcare students practice patient interactions through interactive simulations that provide real-time feedback. Their platform supports expressive text-to-speech, clinical communication evaluation, and emotion-aware understanding to help learners build confidence in patient communication.
Connect with CuraVoice on LinkedIn and Instagram
How CuraVoice Uses PyTorch
PyTorch is the core deep learning engine behind CuraVoice’s voice AI stack. The team uses PyTorch to train expressive text-to-speech models, emotion-aware speech encoders, and a Speech Small Language Model that evaluates empathy and communication quality in real time.
Why CuraVoice Chose PyTorch
CuraVoice selected PyTorch for its flexibility, production reliability, and ecosystem support. The team relies on PyTorch’s dynamic computation graph, NVIDIA NeMo and Riva integrations, and strong tooling for speech and transformer-based models.
What’s Next for CuraVoice
CuraVoice is expanding into nursing, physician assistant, and medical school training programs while collecting more diverse clinical communication data to strengthen its evaluation pipeline. The team is developing multilingual support and building new partnerships with universities, simulation centers, and healthcare systems.
PyTorch Conference 2025 Startup Showcase
Thank you to every startup that participated in the 2025 PyTorch Startup Showcase. Your time, energy, and commitment to building real-world AI applications continue to strengthen the community and inspire the broader ecosystem.
Thank you to our Startup Showcase MC, Chappy Asel, Co-Founder, The GenAI Collective, and to thee judges for their thoughtful evaluation and support including Irving Hsu, Principal, Mayfield; Radhika Malik, Partner, Dell Technologies Capital; Kevin Crosby, Sr. Director, Open Source Funding, GitHub; Denise Teng, Partner, Gradient, and Simon Tiu, Principal, Vertex Ventures US.
Congratulations again to Runhouse and CuraVoice. We look forward to seeing what you build next and to continuing to highlight the companies advancing the future of open source AI.