GPU-Accelerated Quantum-HPC Research Intern

IBM; Town of Yorktown, Northern (NY, KY); Hybrid; USD 34,000 - 55,000; Full time

Job summary

IBM Research seeks interns to accelerate and scale hybrid HPC–quantum applications, focusing on the HPC side: starting from sequential proof-of-concept code, parallelizing and distributing it, and accelerating it with GPUs where possible. You will work with quantum researchers and HPC experts guiding you through the project.

You will analyze code for parallelization, offload portions to GPUs, and run at scale while evaluating performance and scalability, informing how larger problems can be

Qualifications

• Proficiency in Python and modern software engineering practices.
• Proficiency in at least one systems programming language (e.g., C, C++, Rust).
• Hands-on experience with HPC software (OpenMP, MPI, CUDA).
• Ability to communicate technical results clearly to HPC engineers and domain scientists; able to work independently in research.

Responsibilities

• Analyze sequential code to identify parallelization opportunities (shared- and distributed-memory).
• Identify code portions to offload to GPUs for performance gains.
• Run code at scale (hundreds of GPUs) and assess performance and scalability.
• Collaborate with quantum scientists to evaluate impact on research and compare to state-of-the-art classical algorithms.

Skills

Python

C/C++

Rust

CUDA

OpenMP

MPI

GPU offload

Education

Bachelor's degree

Master's degree (preferred)

Tools

OpenMP

MPI

CUDA

Slurm

pybind11

NumPy/SciPy

CuPy

JAX

PyTorch

mpi4py

Dask

Numba

Kokkos

OpenACC

cuBLAS

cuSOLVER

cuTENSOR

MAGMA

NCCL

Job description

IBM Research seeks interns to accelerate and scale hybrid HPC–quantum applications, focusing on the HPC side: starting from sequential proof-of-concept code, parallelizing and distributing it, and accelerating it with GPUs where possible. You will work with quantum researchers and HPC experts guiding you through the project.

You will analyze code for parallelization, offload portions to GPUs, and run at scale while evaluating performance and scalability, informing how larger problems can be