Welcome to the Modelbit platform.

Modelbit is an infrastructure-as-code platform for running your company’s machine learning models in production.

In practice, this includes deployment, staging, hosting, autoscaling, retraining, re-evaluating, drift detection, shadow deployments, and the rest of the ops workflow.

All this is described in code and configuration in your git repo. You then git push your repo Modelbit to make it true in prod.

We invite you to try it out for free.

keyboard_arrow_down

keyboard_arrow_up

Modelbit is most popular with teams with serious production ML infra needs.

Needs like realtime threat detection with vision models; text understanding on batches of 250M+ documents; fraud detection and loan underwriting; and other demanding workloads.

Cody Greco

Co-Founder and CTO

"We retrain and redeploy thousands of models daily, so choosing the right partner for model deployment is critical. We initially investigated SageMaker, but Modelbit’s performance paired with its ease of use was a game-changer for us."

Nick Pilkington

Co-Founder and CTO

"Modelbit enabled us to easily deploy several large vision transformer-based models to environments with GPUs. Rapidly deploying and iterating on these models in Modelbit allowed us to build the next generation of our product in days instead of months."

Daniel Mewes

Staff Software Engineer

"With Modelbit, we’ve been able to experiment with new models faster than ever before. Once a model is up, we’re able to make iterations in a manner of minutes. It feels a bit like magic every time I make a change to our model code, push it to Github, and see it live on Modelbit just seconds later."

keyboard_arrow_down

keyboard_arrow_up

Modelbit's interfaces are git, Python, CLI and web.

Modelbit provides a high-level Python API designed to be used by data scientists and ML researchers from Python notebooks. We also provide a CLI and web application for managing the staging and production environments.

keyboard_arrow_down

keyboard_arrow_up

Each running deployment in Modelbit is containerized and isolated so that customers don’t impact each other.

Running deployments are idempotent. Each git push creates a new version. Each version gets its own container with a unique REST API and Snowflake API. By default, these deployments are run in our cloud; they can be configured to run in yours.

We think these design decisions are why we've achieved 99.99% uptime over a period of years.

keyboard_arrow_down

keyboard_arrow_up

Modelbit is a Tom & Harry joint.

You may know us from Periscope Data, which we founded and grew to $25M-ish ARR before we sold it. Real OGs know us from LAN parties and programming competitions in Rochester, NY circa 2006.

We're venture-funded. Backed by Susa Ventures, Homebrew, SV Angel, and lots more funds and angels. We've got good money from good people, is what we're saying.

You can find us at founders@modelbit.com; in our office in San Francisco trying to find the Goldilocks autoscaling algorithm; and at Warriors games when the same-day tickets are cheap. We hope you'll get in touch.

(terms of service)