Skip to main content

Announcing Transformer Lab for Teams: The Operating System for AI Research Labs

· 2 min read
Tony Salomone
Tony Salomone
Person

Our open source research initiative recently closed a new round of funding and today, we launched the public beta for Transformer Lab for Teams: a modern operating system for AI research labs. It’s open source and free to use.

Over the past year, we spoke with leading AI research labs globally to solve friction in their daily workflows. Most institutional research environments today are fragmented. Researchers often navigate a chaotic landscape of legacy scripts and manual workflows to manage billion dollar compute resources.

Transformer Lab for Teams unifies the entire machine learning lifecycle. It replaces "stitched-together" infrastructure with a single, intuitive interface specifically optimized for the high-velocity requirements of modern research labs.

Built by researchers for researchers to do their best work on the tools they’ve always wanted:

  • Unified Interface: A single dashboard to manage data ingestion, model fine-tuning, and evaluation.
  • Seamless Scaling: The platform is architected to run locally on personal hardware (Apple Silicon, NVIDIA/AMD GPUs) and scale effortlessly to high-performance computing clusters using orchestrators like Slurm and SkyPilot.
  • Open Extensibility: A flexible plugin system allows researchers to add custom training loops, evaluation metrics, and model architectures without leaving the platform.
  • Privacy-First Design: The platform processes data within the user's infrastructure, whether on-premise or in a private cloud, ensuring sensitive research data never leaves the lab's control.
  • Complexity Made Simple: Capabilities that used to require complex engineering are now built-in.
    • Capturing checkpoints (with auto-restart)
    • One-line to add hyperparameter sweeps
    • Storing artifacts in a global object store accessible even after ephemeral nodes terminate.

Ultimately, all results are automatically tracked, visualized, and sharable, so you don’t have to wrestle with the logistics of your data. Our long-term goal is to advance and democratize the essential tools used by every AI research lab in the world.

If you’re an AI/ML researcher looking to improve how you work, join the beta here.