We’re partnering with Sunscreen to make TFHE practical for real SaaS applications. Sunscreen brings the Parasol compiler for TFHE programs; Lattica brings the runtime, key lifecycle, and deployment surface that turn those programs into live encrypted services.
With Sunscreen’s Parasol compiler on the program side and Lattica’s runtime, key lifecycle, and deployment surface on the operations side, teams can take a TFHE program from source to a live encrypted SaaS endpoint without assembling that stack themselves.
Why TFHE, and where it fits alongside CKKS
FHE is not one thing. The right scheme depends on what your program actually needs to do on encrypted data. CKKS is the workhorse for AI and data-heavy workloads: approximate arithmetic, large parallel batches, a natural fit for GPU acceleration. TFHE is the right tool when your logic branches on encrypted values: comparisons, conditionals, policy checks, exact integer logic. Different shape of computation, different scheme. Lattica supports both.
CKKS
Approximate arithmetic on packed vectors of values. Built for AI inference, analytics, and any workload dominated by linear algebra. GPU-accelerated through Lattica’s stack.
TFHE
Exact computation on encrypted bits and integers, with bootstrapping built into every gate. The right scheme for conditional logic, voting, smart-contract state, and policy evaluation.
How the partnership works: Sunscreen compiles, Lattica runs
The flow is straightforward: a developer compiles a TFHE program with Sunscreen’s Parasol, then hands the artifact to Lattica, which runs it as a queryable encrypted service.
- Create a key set
- Encrypt
- Decrypt
- Deploy evaluation key
- Encrypted query
- Encrypted response
Lattica Sunscreen
-
Compile with Sunscreen
Write your TFHE program and compile it locally with Parasol. The output is a deployable encrypted artifact, not a research notebook.
-
Deploy to Lattica
Upload the artifact. Lattica provisions the runtime, allocates compute, and exposes the program as a queryable encrypted endpoint.
-
Configure access and resources
Set permissions for who can issue queries, choose the compute tier, and control credit usage from a single console.
-
Run encrypted queries
Clients use Lattica’s management library to generate keys, encrypt inputs, send queries, and decrypt encrypted responses. The server never sees plaintext.
Where TFHE fits
TFHE is the right scheme when your application requires exact computation over encrypted inputs, not approximate, not statistical.
Smart contracts
Exact balance checks and conditional logic over encrypted wallet state, without revealing inputs to the chain.
Access control and policy engines
Evaluating rich conditions over private user attributes without exposing either the policy or the data being checked.
Decision trees
Traverse encrypted decision trees end to end, returning a single encrypted leaf without leaking the path, the splits, or the input features.
Sealed-bid auctions
Compare encrypted bids and resolve the winner under exact rules, with neither the auctioneer nor the participants ever seeing competing bids in the clear.
What this unlocks
The combination of a real TFHE compiler and a managed runtime is what makes encrypted SaaS practical. Teams stop building cryptographic infrastructure and start shipping the application logic that’s actually theirs.
“TFHE has come a long way over the past few years. Between advances in the underlying schemes and the work the Sunscreen team has put into the compiler, the primitives are in a very different place than they were even two years ago. The missing piece has been the operational layer around them. With Sunscreen on the compiler side and Lattica on the runtime side, that layer finally exists.”
- Lattica engineering team
Get started
Compile a TFHE program with Sunscreen, deploy it on Lattica, and run encrypted queries against it from your own client. New accounts get free compute hours so you can take a real workload end to end before committing.
