Lead GPU Programming Engineer

We are looking for a lead GPU programming engineer to optimize our GPU-accelerated proving system. You will work with the crypto team to implement optimized CUDA code for the crypto primitives underlying modern zero-knowledge proofs. This work is critical for increasing the throughput and decentralization of our fully EVM-compatible zkRollup. Your day-to-day work will involve proposing and implementing optimizations to our current prover as well as packaging the system in an easy-to-use SDK. Our existing system targets the UltraPlonk arithmetization using KZG polynomial commitments and already achieves a 10x+ speed improvement over the state-of-the-art on core computations. The ideal candidate will have prior experience implementing efficient cryptographic code, a deep understanding of GPU acceleration, or a deep understanding of zero-knowledge proof protocols.

Job Duties

• Design and implement optimization to GPU kernels for prover algorithms
• Profile and identify bottlenecks in GPU kernel performance
• Integrate GPU-accelerated kernels into our Rust prover.
• Optimize the pipeline of memory transfer and kernel execution in the prover
• Implement easy-to-use interfaces to our GPU prover

Qualifications

• Proficiency in CUDA.
• Experience implementing and optimizing accelerated kernels on GPUs
• Fluency in Rust, C++, or similar languages. Rust is preferred, but prior experience is not mandatory
• Prior experience optimizing low-level cryptography code is preferred.
• Strong background in math, cryptography, or related areas is preferred.

Benefits

• This is a remote position. Work from anywhere with our global team.