Description
As a Senior Systems Software Engineer focused on performance architecture for GPU-accelerated structured data processing, you will extend JIT and compiler-based execution support in cuDF and related GPU-accelerated structured data processing systems.
Design approaches for lowering expressions, ASTs, or query fragments into optimized GPU execution paths.
Investigate kernel fusion strategies across cuDF operations to reduce materialization, memory traffic, launch overhead, and end-to-end query latency.
Analyse structured analytics workloads to identify performance bottlenecks in expression evaluation, joins, aggregations, scans, data movement, and memory management.
Build benchmarks and regression tests that capture real dataframe and SQL-like workloads, from micro-benchmarks to end-to-end pipelines.
Collaborate with cuDF, CUDA, compiler/runtime, and query engine teams to translate workload analysis into implementation plans and architecture decisions.
Prototype and evaluate execution strategies inspired by high-performance database engines, including fused operators, code generation, vectorized execution, and adaptive planning.
This is a high-impact individual contributor role for someone passionate about developing coordinated SQL and user-friendly interfaces across diverse CPU and GPU query engines.
The ideal candidate has deep experience in systems performance, compiler/runtime design, and database or dataframe execution engines.
This role will focus on compiler and JIT-based execution techniques for cuDF and related analytics runtimes.
The ideal candidate has deep experience in systems performance, compiler/runtime design, and database or dataframe execution engines.
Proven skills in profiling, instrumentation, and optimisation for CPU and GPU systems, applying tools like tracing, counters, flame graphs, and kernel-level profiling.
Experience with compiler, JIT, code generation, query execution, or runtime optimisation techniques.
Experience optimising analytic database engines and/or query runtimes, including vectorised execution, join strategies, and columnar formats like Arrow and Parquet.
Proficient in C++ and/or Python, with a strong ability to analyse performance-critical code and implement effective solutions.
Experience with cuDF, RAPIDS, CUDA, Numba, LLVM, MLIR, NVRTC, or other JIT/codegen systems.
Experience with benchmarking frameworks, performance dashboards, and CI/CD regression gating, along with a proven grasp of modern analytics and machine learning workflows.
Deep familiarity with NVIDIA GPUs and GPU programming (CUDA), including memory hierarchy, concurrency, and profiling toolchains such as Nsight Systems.
Experience with TPC-style benchmarking (TPC-H, TPC-DS, or analogous), Click-Bench-like workloads, and building credible, repeatable performance narratives.
Prior work on database execution engines, especially operator fusion, query compilation, vectorised execution, or adaptive execution.
Demonstrated open-source contributions to performance-critical systems, including libraries, runtimes, databases, and ML or data tooling.
With a competitive salary package and benefits, NVIDIA is widely considered to be one of the technology world’s most desirable employers.
We have some of the most forward-thinking and hardworking people in the world working for us.
Are you a creative and autonomous Systems Software Engineer, who loves challenges?
Do you have a genuine passion for advancing the state of Data Intelligence across a variety of industries?
If so, we want to hear from you.
You will also be eligible for equity and benefits.
Applications for this job will be accepted at least until June 16, 2026.
This posting is for an existing vacancy.
NVIDIA uses AI tools in its recruiting processes.