Type: Full-Time

Join a collaborative research effort with a team of world-class scientists to shape the next generation of high-performance computing tools for large-scale optimization. We're looking for exceptional researchers to work on foundational operators akin to BLAS, SparseBLAS, GraphBLAS, LAPACK, and similar libraries. The goal is to develop core computational building blocks that power advanced optimization workloads.

Key Responsibilities

As a Research Scientist, you will:

• Identify essential and emerging basic operations relevant to high-performance optimization platforms.
• Conduct in-depth "speed-of-light” analyses to uncover:
• Performance bottlenecks,
• Scalability characteristics (e.g., iso-efficiency),
• Trade-offs (e.g., memory vs. communication),
• Optimal device configurations (e.g., mix of CPUs and accelerators).
• Design and prototype high-performance, scalable, and productive software systems that serve as the computational foundation for modern optimization solvers.

You’ll be expected to:

• Design and implement novel operator-level routines tailored for optimization tasks.
• Analyze algorithms for their theoretical and practical performance under parallel computation models, considering computation, memory access, and data reuse.
• Apply cache-aware or cache-oblivious techniques and HPC best practices (shared/distributed-memory parallelization, vectorization).
• Research and design data structures for acceleration on CPUs and other processing units (e.g., AI accelerators, GPUs).
• Enable solvers to be written in a modular, data-centric style with clear and efficient control flows.
• Contribute to or extend existing run-time systems and communication frameworks to improve performance, scalability, and automation of trade-offs.
• Guarantee robust performance and correctness of solvers built atop these operators, with applications in science and industry.

Required Qualifications

Candidates should bring solid experience in several of the following areas:

• Optimization of irregular algorithms, such as graph computations or sparse linear algebra, from high-level algorithm design to low-level optimizations like SIMD or locking strategies.
• Multi-core or many-core programming using technologies like POSIX Threads, OpenMP, or similar.
• Distributed-memory computing (e.g., MPI, BSP), including experience with collective communications or RDMA.
• Experience with performance-tuned code generation frameworks (e.g., ALP, BLIS, DaCE, Spiral, FLAME, Firedrake).
• Strong C++ (C++11 or later) skills, particularly in generic programming, algorithms, and data structures.
• Proficiency with debugging and performance analysis tools (e.g., Valgrind, GDB, CI systems).
• A strong publication record in top-tier HPC or applied mathematics venues.
• Excellent communication skills with the ability to clearly convey complex technical content.
• A collaborative mindset and comfort working in diverse, international teams.

Preferred Qualifications

Experience with the following is considered a plus:

• GraphBLAS or Algebraic Programming paradigms.
• Optimization methods or solver design.
• High-performance interconnects and programming (e.g., Infiniband, RDMA).
• Accelerator programming (e.g., CUDA, OpenCL).
• Publications in physical sciences or theoretical computer science.

If interested please reach out to leyton@microtech-global.com