High-Performance & Scientific Computing — Skills Plan
Planned skill map for High-Performance & Scientific Computing. The curriculum will be developed and released in stages.
Measure and optimize numerical workloads with parallel algorithms, distributed computing, GPU acceleration, profiling, reproducibility, and error analysis.
Potential roles: HPC Engineer · Scientific Software Engineer · GPU Computing Engineer
Before you start
- C/C++ or Python programming
- Linear algebra and numerical methods
- Linux, shell, and computer architecture
What you’ll build
- Parallelize and benchmark a numerical workload
- Profile a GPU-accelerated computation
- Reproduce and validate a scientific result
Prerequisites
PLANNED- C/C++ or Python programming
Build the C/C++ or Python programming knowledge needed to follow the High-Performance & Scientific Computing foundations.
- Linear algebra and numerical methods
Build the Linear algebra and numerical methods knowledge needed to follow the High-Performance & Scientific Computing foundations.
- Linux, shell, and computer architecture
Build the Linux, shell, and computer architecture knowledge needed to follow the High-Performance & Scientific Computing foundations.
Foundations
PLANNED- Performance measurement and algorithmic cost
Explain how Performance measurement and algorithmic cost works and where it fits in the High-Performance & Scientific Computing learning path.
- Floating-point representation and numerical error
Explain how Floating-point representation and numerical error works and where it fits in the High-Performance & Scientific Computing learning path.
- Memory hierarchy and vectorization
Explain how Memory hierarchy and vectorization works and where it fits in the High-Performance & Scientific Computing learning path.
- Parallelism and concurrency fundamentals
Explain how Parallelism and concurrency fundamentals works and where it fits in the High-Performance & Scientific Computing learning path.
Core Skills
PLANNED- Multithreading and distributed-memory concepts
Use Multithreading and distributed-memory concepts to implement a working capability and verify its behavior in High-Performance & Scientific Computing.
- GPU programming and accelerator models
Use GPU programming and accelerator models to implement a working capability and verify its behavior in High-Performance & Scientific Computing.
- MPI and scientific workload orchestration
Use MPI and scientific workload orchestration to implement a working capability and verify its behavior in High-Performance & Scientific Computing.
- Profiling, benchmarking, and reproducibility
Use Profiling, benchmarking, and reproducibility to implement a working capability and verify its behavior in High-Performance & Scientific Computing.
Advanced / Production
PLANNED- Load balancing and communication costs
Evaluate Load balancing and communication costs trade-offs against the performance, reliability, and safety needs of High-Performance & Scientific Computing systems.
- Numerical stability and validation
Evaluate Numerical stability and validation trade-offs against the performance, reliability, and safety needs of High-Performance & Scientific Computing systems.
- Cluster scheduling and resource efficiency
Evaluate Cluster scheduling and resource efficiency trade-offs against the performance, reliability, and safety needs of High-Performance & Scientific Computing systems.
- Scalable data and checkpointing strategies
Evaluate Scalable data and checkpointing strategies trade-offs against the performance, reliability, and safety needs of High-Performance & Scientific Computing systems.
- Quantum computing concepts and limitations
Evaluate Quantum computing concepts and limitations trade-offs against the performance, reliability, and safety needs of High-Performance & Scientific Computing systems.
Real-World Projects
PLANNED- Parallelize and benchmark a numerical workload
Deliver a working outcome with Parallelize and benchmark a numerical workload, then test it against the project requirements.
- Profile a GPU-accelerated computation
Deliver a working outcome with Profile a GPU-accelerated computation, then test it against the project requirements.
- Reproduce and validate a scientific result
Deliver a working outcome with Reproduce and validate a scientific result, then test it against the project requirements.
Career / Interview Preparation
PLANNED- Parallel performance analysis
Explain your approach to Parallel performance analysis, defend key trade-offs, and discuss what you learned.
- Numerical correctness review
Explain your approach to Numerical correctness review, defend key trade-offs, and discuss what you learned.
- HPC project and reproducibility walkthrough
Explain your approach to HPC project and reproducibility walkthrough, defend key trade-offs, and discuss what you learned.
Discussion
0Have a better approach? Found something outdated? Share it — your knowledge helps everyone learning here.