get early access see demo
get early access see demo

Heal:

A Standard Interface for Accelerated FHE Computing

HEAL bridges the gap between FHE software and hardware, enabling integration of acceleration hardware with Lattica’s FHE-as-a-Service platform

become a partner

What is HEAL

HEAL (Homomorphic Encryption Abstraction Layer) is a hardware-
friendly interface designed to standardize FHE acceleration.
It abstracts away the complexity of FHE software, enabling integration with any FHE hardware accelerator.

HEAL in the FHE Stack
FHE Applications
Encrypted DB AI Inference Smart Contracts
FHE Software Stack
Algorithms
BGV CKKS TFHE GSW
Core Building Blocks
Polynomial
arithmetic KeySwitch Bootstrap
HEAL

Universal Interface
for FHE Operations

img img img img img img img
Acceleration Hardware

host_to_device

compute instructions

device_to_host

img img

FPGA

img img

ASIC

img img

TPU

img img

GPU

why hardware vendors choose HEAL

Focus on Hardware, Stay Aligned with Software

Ensures compatibility with the latest FHE software stack, allowing hardware vendors to optimize performance without needing to track ongoing software changes.

FHE Without Cryptographic Complexity

Abstracts away the
complexities of homomorphic encryption, allowing hardware teams to integrate FHE support without requiring deep cryptographic expertise.

Built for Future FHE Applications

Designed to support emerging workloads like AI inference and secure analytics, enabling hardware to integrate seamlessly as the field evolves.

Access to a Growing FHE Market

Connects hardware vendors with enterprises and researchers using Lattica’s platform, accelerating adoption in real-world encrypted computation.

How to Get Started

Join the HEAL Partner Program

We are partnering with leading FHE hardware vendors to define the next generation of secure computation.

    FAQ

    HEAL is designed to be hardware-agnostic, supporting a wide range of FHE acceleration architectures, including:

    • Custom ASICs optimized for FHE
    • FPGAs with reconfigurable logic for FHE
    • GPUs and TPUs leveraging parallelism for modular arithmetic over tensors

    HEAL provides a standardized interface for FHE accelerators, handling the allocation and management of memory and efficiently dispatching computation to hardware. Integration happens through the following key mechanisms:

    • Memory Management & Data Flow Control
      Handles memory allocation and movement, ensuring efficient use of hardware resources via optimizations in the algorithmic layer.
    • Instruction Dispatch & Execution
      Translates high-level FHE operations into optimized instruction sequences for the accelerator, abstracting away low-level cryptographic details.
    • Hardware-Specific Optimizations
      Allows for vendor-specific extensions, ensuring accelerators can leverage their unique architectures while remaining compatible with standard FHE frameworks.

    To integrate HEAL with your accelerator, we provide detailed API documentation, a reference implementation, and direct collaboration support for hardware vendors.

    Lattica’s FHE-as-a-Service platform is a privacy-preserving AI and computing solution that allows enterprises to run encrypted workloads without exposing sensitive data.

    HEAL is a core technology within this platform, ensuring that FHE workloads run efficiently on specialized hardware. By supporting HEAL, your hardware becomes compatible with Lattica’s platform, making it easier for enterprise customers to adopt your acceleration technology.

    Yes! We actively collaborate with hardware vendors, researchers, and industry experts to refine HEAL.
    Ways to contribute:

    • Provide feedback on performance optimizations and hardware-specific enhancements.
    • Propose extensions to support new FHE primitives or architectures.
    • Work with our team to ensure compatibility with your hardware.

    Interested in contributing? Contact us to discuss partnership opportunities.