SF Tensor YC Application
SF Tensor went through Y Combinator in the Fall 2025 batch as a B2B company. Below: what the company does, who founded it, and its application to YC, reconstructed question by question in the real YC form.
- YC batch
- Fall 2025
- Industry
- B2BInfrastructure
- Team size
- 6
- Status
- Active
- Launched
- 2025
Founders
Ben KoskaFounder
Luk KoskaFounder
Tom KoskaFounder
What SF Tensor does
AI researchers should be pushing the boundaries of what's possible with new architectures and training methods. Instead, they waste weeks configuring cloud infrastructure, debugging distributed systems, and optimizing their GPU code. We know because we lived it: While training our own models across thousands of GPUs earlier this year, we spent more time fighting our infrastructure than doing actual research. That's why we're building two things. First, Elastic Cloud: a managed platform that automatically finds the cheapest GPUs across all providers, handles spot instance preemption, and cuts compute costs by up to 80%. Second, automatic kernel optimization that makes training code run faster by modeling hardware topology, often beating hand-tuned implementations. The problem is that getting high performance across different hardware is genuinely hard. NVIDIA's CUDA moat exists because writing fast kernels requires deep expertise. Most teams either accept vendor lock-in or hire expensive kernel engineers. Our goal is to break the CUDA moat. The compute bottleneck is the biggest constraint on AI progress. NVIDIA can't manufacture enough GPUs, and their monopoly keeps prices astronomical. Meanwhile, AMD, Google, and Amazon are shipping capable alternative hardware that nobody uses because the software is too hard. We're breaking that moat. If we succeed, anyone will be able to train state-of-the-art models without thinking past their PyTorch code.
SF Tensor's YC application, reconstructed
Y Combinator application
SF Tensor, Fall 2025 batch. The real form questions, answered from public data.
- Founders2
- Company7
- Progress7
- Idea4
- Equity3
- Curious2
- Batch Preference1
Company
SF Tensor
Infrastructure for AI labs to focus on research
We're building two core products. First, Elastic Cloud, a managed platform that automatically finds the cheapest GPUs across cloud providers, handles spot instance interruptions, and cuts compute costs by up to 80%. Second, automatic kernel optimization that models hardware topology to speed up training code, often surpassing hand-tuned implementations. Our goal is to remove infrastructure friction so AI researchers can spend more time experimenting and less time on system ops.
We do not have a set company location; our team is distributed. We expect to remain remote after YC.
Because our work is mostly software infrastructure targeting global AI labs, we do not depend on location. Remote lets us access talent and collaborate with AI researchers worldwide.
21 more answers in the full application
Founders, Progress, Idea, Equity, Curious, Batch Preference are in the full version. Create a free account to open it in YC Insights, next to every other YC batch.
Frequently Asked Questions
Is this the real SF Tensor YC application?
No. Y Combinator keeps submitted applications private. This is a reconstruction: the real YC form questions, answered from SF Tensor's public data (launch notes, founder backgrounds, product description). It shows how a company like SF Tensor would present itself to YC, not the text they submitted.
What YC batch was SF Tensor in?
SF Tensor was part of the Y Combinator Fall 2025 batch.
Who founded SF Tensor?
SF Tensor was founded by Ben Koska, Luk Koska, Tom Koska.
What does SF Tensor do?
Infrastructure for AI labs to focus on research.
How do I read the full reconstructed application?
Create a free Round Funded account. The full 26-answer version of SF Tensor's application, and every other YC batch, is in YC Insights.
