As an R&D Intern, you will be fully integrated into the engineering team. You will work on concrete features of our compiler and help us validate our technology against real-world scenarios.
Think: Analyze the performance of generated code. You will study how specific algorithms behave on a GPU versus a CPU and identify bottlenecks. You will learn to read and understand the "Assembly" (PTX/LLVM) generated by our tools.
Implement:
Develop new intrinsic functions to expose specific hardware capabilities (like Tensor Cores or SIMD instructions) to high-level languages.
Port standard mathematical algorithms (linear algebra, stencils, image processing) using Hybridizer to demonstrate our capabilities.
Fix bugs and improve the robustness of the compiler infrastructure.
Build:
Benchmarking & Testing: Create automated performance tests to compare our generated code against hand-written CUDA/C++.
Samples & Demos: Build simple, clean code examples and "Getting Started" projects to help new users understand how to use Hybridizer.
Documentation: document new features and create technical content explaining how to optimize code using our tools.
We are looking for a student passionate about low-level programming and performance. You don't need to know everything yet, but you must be eager to learn complex concepts.
Education: Currently enrolled in a French "Grande École" d'Ingénieurs or a Master 2 in Computer Science (specialization in HPC, Systems, or Applied Math is a plus).
Core Coding Skills: Strong proficiency in C++ or C#. You understand pointers, memory management, and object-oriented programming.
HPC Interest: You have some knowledge of (or a strong desire to learn) CUDA, OpenMP, or GPU architecture.
Math & Logic: You are comfortable with algorithmics and standard engineering mathematics (linear algebra, numerical methods).
Mindset: You are rigorous, curious, and comfortable reading technical documentation in English. You enjoy solving puzzles and understanding how things work "under the hood."
Skills Required
- Currently enrolled in a French Grande École d'Ingénieurs or Master 2 in Computer Science
- Strong proficiency in C++ or C#
- Understanding of pointers, memory management, and object-oriented programming
- Knowledge of, or strong desire to learn, CUDA, OpenMP, or GPU architecture
- Comfort with algorithmics, linear algebra, and numerical methods
- Ability to read technical documentation in English
- Specialization or interest in high-performance computing, systems, or applied mathematics
What We Do
Hybridizer is a software platform for performance portability and GPU acceleration. Its compiler transforms C#/.NET and Java bytecode or high-level code into optimized source code for multicore CPUs and GPUs, allowing developers to use existing codebases without learning CUDA or rewriting applications. The technology supports debugging, profiling, cross-platform deployment, and demanding workloads such as quantitative finance, scientific simulation, and data processing.
