Tian Yu (T.Y.)

Were building the worlds most automated AI lab. Our objective is to build systems that optimize and automate work, starting with research itself. We dont think the next step change in AI will come from scaling the current recipe: larger models, more data, static deployment. Were pursuing new learning algorithms that supersede large-scale pretraining and reinforcement learning, and architectures that scale better than transformers. — We think the next wave of frontier research will come from small teams with highly capable agents. Were building the lab around that bet from day one. We start by automating our own work. That makes room for more ambitious and creative work, which reveals the next thing to automate. What we learn feeds back into the research. — If were right, small teams with powerful AI systems will take on work that once required entire organizations. Much of the current conversation around AI assumes it will mostly help existing institutions operate more efficiently. Were excited by the larger shift: a world where far more people can pursue ambitious work without first building an organization or raising capital. A corner store with the logistical capabilities of a multinational corporation. A small workshop with demand from around the world. A project that starts in a group chat and grows far beyond its origins. Were building the research, systems, and operating model to make this possible for more people. — Our team brings together people who have helped build frontier models, state-of-the-art optimization methods, influential architectures, multimodal agents, and the low-level systems those advances depend on. Were building a small lab with the range to rethink neural network architecture from scratch – the kind of bet that gets harder to make the more you have to scale – and the systems engineering depth to prove it out.

Recursive self-improving superintelligence to automate knowledge discovery.

The next-generation AI developer platform unifying the development and deployment of AI for the world.

BentoML is an enterprise-grade Inference platform for deploying and managing AI models at scale. It offers full control without the complexity, allowing teams to serve any model including LLMs, embeddings, and agentic pipelines across VPC, on-prem, or hybrid environments with tailored optimization, advanced orchestration, and fine-grained performance tuning. From prototype to production, BentoML covers the full inference lifecycle with instant model deployments, elastic autoscaling, built-in observability, compliance-ready features, and mission-critical reliability, freeing your team to deliver AI that drives real business outcomes faster.

Were building the worlds most automated AI lab. Our objective is to build systems that optimize and automate work, starting with research itself. We dont think the next step change in AI will come from scaling the current recipe: larger models, more data, static deployment. Were pursuing new learning algorithms that supersede large-scale pretraining and reinforcement learning, and architectures that scale better than transformers. — We think the next wave of frontier research will come from small teams with highly capable agents. Were building the lab around that bet from day one. We start by automating our own work. That makes room for more ambitious and creative work, which reveals the next thing to automate. What we learn feeds back into the research. — If were right, small teams with powerful AI systems will take on work that once required entire organizations. Much of the current conversation around AI assumes it will mostly help existing institutions operate more efficiently. Were excited by the larger shift: a world where far more people can pursue ambitious work without first building an organization or raising capital. A corner store with the logistical capabilities of a multinational corporation. A small workshop with demand from around the world. A project that starts in a group chat and grows far beyond its origins. Were building the research, systems, and operating model to make this possible for more people. — Our team brings together people who have helped build frontier models, state-of-the-art optimization methods, influential architectures, multimodal agents, and the low-level systems those advances depend on. Were building a small lab with the range to rethink neural network architecture from scratch – the kind of bet that gets harder to make the more you have to scale – and the systems engineering depth to prove it out.

Recursive self-improving superintelligence to automate knowledge discovery.

The next-generation AI developer platform unifying the development and deployment of AI for the world.

BentoML is an enterprise-grade Inference platform for deploying and managing AI models at scale. It offers full control without the complexity, allowing teams to serve any model including LLMs, embeddings, and agentic pipelines across VPC, on-prem, or hybrid environments with tailored optimization, advanced orchestration, and fine-grained performance tuning. From prototype to production, BentoML covers the full inference lifecycle with instant model deployments, elastic autoscaling, built-in observability, compliance-ready features, and mission-critical reliability, freeing your team to deliver AI that drives real business outcomes faster.