I am Cristian Ramírez, a performance-focused R&D engineer working on generative AI, with a PhD in Computer Science. My work connects AI inference, low-level kernels, runtime optimization and computer architecture.
Current engineering work
My current responsibilities include optimizing compute kernels and AI runtimes, building agentic systems for performance optimization, and benchmarking and profiling AI workloads. I also contribute to software development, testing, code review and technical documentation.
I work with AI agents and vLLM for inference, alongside CUDA, C++ and Clang. My engineering focus is the interaction between the inference software stack and the hardware that executes it.
I am looking for engineering roles in AI inference, performance optimization, GPU computing and computer architecture. I want to contribute to teams that implement, profile and improve the systems behind AI workloads.
Research and implementation
From January 2021 to April 2024, I worked as a Research Engineer at Universitat Politècnica de València (UPV), developing and optimizing linear algebra kernels for deep learning inference on low-power ARM and RISC-V processors.
As a Research Assistant & Engineer, I contributed to work supported by eFlows4HPC (Horizon 2020 / EuroHPC), the TED2021-129334B-I00 project with EU NextGenerationEU–PRTR support, and the Spanish PID2020-113656RB-C22 project. The research project details connect these programmes to the publications and their funding acknowledgements.
I focus on efficient computing for numerical and machine learning applications. My research covers matrix multiplication, deep learning convolutions and their implementation on low-power RISC-V processors. The work considers both arithmetic and the cost of moving data through memory hierarchies.
The implementation described in my GAP8 article uses BLIS-like int8 GEMM on the Fabric Controller, with packing and explicit data movement through the memory hierarchy. The SDK used in the published evaluation emulated the vector dot-product operation in software.
Education
- PhD in Computer Science (2020–2025). Universitat Politècnica de València (UPV). Thesis: RISC-V Low-Power Processors: Enhancing Matrix Computations and Deep Learning Convolution Algorithms.
- MSc in Electronic Systems Engineering — Digital Systems (2018–2019). Universitat Politècnica de València (UPV).
- BSc in Telecommunications and Electronics Engineering (2010–2015). UTPL.
The publication list links to the institutional records and preserves the original titles.
Engineering skills and interests
My research engineering experience includes ARM NEON and RISC-V Vector (RVV) intrinsics, OpenMP multithreading, and RTOS integration and validation under real-time and memory constraints. I use Python and Bash to automate correctness checks, benchmarking and performance regression tracking.
My current interests include inference performance, GPU computing, compiler behavior and agent systems. My published work adds depth in numerical computing, memory hierarchies, RISC-V, embedded AI and hardware–software co-design.
I am interested in engineering opportunities where implementation, profiling and architectural understanding are central to the work. Contact me.