Deploying TFHE in Production with Lattica

At Sunscreen, we've spent the past few years building what we believe is the most developer-friendly TFHE compiler out there. Our Parasol compiler lets you write your program in C, add a few directives to indicate which inputs should be encrypted, and we handle the rest - parameter selection, circuit optimization, bootstrapping, all of it. If you've read our previous posts, you know how strongly we feel about letting developers "bring their own program" rather than rewriting everything from scratch.

But there's always been a gap between having a working TFHE program and running it in production.

The missing piece

Let's be honest about the state of things. You can write an FHE program, compile it, and run it locally. But what happens when you actually want to deploy it? Suddenly you're managing compute infrastructure, figuring out key lifecycle management, handling orchestration, and worrying about resource allocation. These are hard problems, and they're not cryptography problems. Most teams building with FHE are not (and shouldn't have to be) infrastructure teams.

This is the gap we've been thinking about for a while: how do we get from "cool demo" to "production service"?

Lattica

Today we're excited to announce our partnership with Lattica to bring practical TFHE to production environments.

Here's how the two pieces fit together:

Sunscreen provides the cryptographic engine via our Parasol TFHE compiler transforming your program into optimized encrypted computation, and our runtime executes it. 

Lattica provides the rest of the infrastructure around it: orchestration and workflow automation, compute resource allocation, key management, and monitoring. 

The result is a workflow that feels much closer to deploying a normal application:

1. Compile with Sunscreen: write your TFHE program and compile it locally with Parasol to produce a deployable artifact.
2. Deploy to Lattica: upload your compiled artifact to Lattica's platform.
3. Configure access and resources: set permissions, choose your compute resources, and control credit usage.
4. Accept encrypted queries: your program is live and ready for use.

If you've ever dealt with the pain of trying to stand up FHE infrastructure from scratch, you'll appreciate how much complexity this abstracts away.

“Sunscreen has built a powerful TFHE compiler that makes encrypted programs accessible to developers. At Lattica, our focus is making those programs deployable, scalable, and operable in real environments. This partnership is about closing the gap between cryptographic capability and production reality, so teams can move from prototype to live encrypted services without reinventing infrastructure,” shared Rotem Tsabary, Founder & CEO at Lattica.

“This partnership represents something we've been working toward since we started Sunscreen. We always believed that TFHE would only see real adoption when deploying it felt as natural as deploying any other application. Lattica brings the infrastructure expertise that complements our cryptographic engine, helping privacy builders turn working demos into production services. We’re excited to see what developers will build,” said Ravital Solomon, CEO at Sunscreen.

Why TFHE (and when not)

If you've been following FHE developments, you know there's more than one FHE scheme out there, and the choice matters quite a bit depending on your application. Lattica’s platform makes encrypted compute a cloud-native capability for CKKS, and provides the execution layer that brings sensitive workloads and cloud infrastructure together, so it's worth briefly touching on when you'd reach for each.

CKKS excels at approximate arithmetic on large datasets. It's parallelizable, GPU-accelerated, and ideal for AI workloads. If you're doing data-heavy batch operations, CKKS is likely your best bet.

TFHE is different. It's built for exact computation: comparisons, conditionals, branching logic. The tradeoff is that it's CPU-based and generally more expensive per operation than CKKS for pure arithmetic.

So when do you want TFHE? When precision is a requirement. Smart contracts that need exact balance checks. Voting systems where rounding errors would be catastrophic. Any application involving complex decision logic over encrypted data - if-else branches, comparisons, lookups - this is where TFHE shines.

Our partnership with Lattica means you don't have to choose between CKKS and TFHE at the infrastructure level. Both are available through the same platform, so you can use the right tool for each part of your application.

What we're excited about

This partnership represents something we've been working toward for a long time: making TFHE not just possible but practical. We've always believed that one of the biggest barrier to adoption isn't the cryptography itself - it's the infrastructure around it like the deployment complexity, the operational overhead etc. By partnering with Lattica, we can focus on what we do best (the cryptographic engine) while Lattica handles the infrastructure challenges that have historically kept TFHE out of production environments.

We think this is how TFHE gets adopted in the real world - not by asking every team to become infrastructure experts, but by making encrypted computation something you can deploy as easily as any other service.

Get started

Lattica is offering 3 free compute hours to get you up and running. You can check out the full workflow and sign up on Lattica's website. For more info on Sunscreen's TFHE compiler and some demos of what's possible, take a look at our documentation and demos.

We're excited to see what you build with this. If you're working on an application that could benefit from encrypted computation - whether it's in finance, healthcare, blockchain, or something we haven't thought of yet - we'd love to hear from you.