Hey YC! We’re Luis and Aamir, founders of Understudy.

TL;DR

We help you cut your company’s Anthropic bill by 80% with no loss of performance.

Understudy is an inference cloud. We capture your employee’s AI work traces, train smaller models, and automatically deploy them (only when they outperform the expensive model you are currently paying for).

Instead of renting the same frontier model forever, your AI infrastructure gets cheaper and better every time it runs.

See more: https://youtu.be/YhfoOchpYaw

The problem

Last week, we met with a company spending $40M/year on Anthropic. And it’s only going in one direction.

To them, it feels totally out of control.

They’re not alone - every company we meet is spending record amounts on models from frontier labs.

But almost none of that spending compounds.

I.e. You send a prompt, pay for the answer, and throw away the experience. Six months later, you are still renting the same general-purpose model at roughly the same price.

Meanwhile, training custom models requires an entire ML team to collect and clean traces, build evals, choose training data, fine-tune and reinforce models, test serving configurations, compare cost, latency, and quality, safely deploy or roll back the winner.

For most companies, it’s just too much to bother with.

What Understudy does

Point Understudy at an AI workload.

We capture how your existing agent performs, turn those traces into evals, and start training smaller models that can do the same job better.

Understudy handles the entire loop:

Observe → evaluate → train → test → deploy → repeat

A new model is promoted only when it beats the incumbent on the metrics you care about, such as:

• task success
• inference cost
• latency
• privacy
• deployment size

Your customized models are owned by you, and can run in our cloud, your cloud, or on-premise.

An example

Suppose your product uses Claude for a repetitive legal workflow.

Understudy observes successful production runs, learns the specific tools and judgment required for that workflow, and trains a specialized open-weight model.

Once that model matches or beats Claude on a held-out evaluation, Understudy starts routing production traffic to it.

As more work is completed, the model keeps improving.

Why us

Aamir and I met building Instacart’s optimization and experimentation systems, growing multiple business lines to more than $1B in annual revenue.

We started Understudy because we believe every company using AI should be accumulating intelligence, not accumulating API bills.

Our ask

Spending too much on repetitive model workloads? Ready to own your own models? We want to meet you - come to understudylabs.com today.