Job Description
Ricursive Intelligence is a frontier AI lab building self-improving systems, starting with chip design. We are reinventing chip development and closing the loop between AI and the hardware that fuels it, recursively accelerating the path to artificial superintelligence. Backed by $335M from Sequoia, Lightspeed, DST, and NVIDIA Ventures, we are a growing, fast-paced team where every hire shapes the work.
The company has unmatched talent density, including IMO, IPhO, and IOAA gold medalists, pioneers who made prior breakthroughs in chip design: AlphaChip (Nature 2021), ePlace (DAC Best Paper Nominee 2014), RL-CCD (DAC Best Paper 2023), INSTA (DAC Best Paper 2025), and C3PO (ASP-DAC Best Paper 2026), chip leads for Apple Silicon (multiple generations of iPhone and iPad) and Google (TPU, OpenTitan), and top researchers and engineers from Anthropic, Google DeepMind, Stanford, and MIT.
ABOUT THE ROLE
Every chip Ricursive designs is only as good as the software that targets it. The compiler is where architecture meets workload: it turns new vector units, matrix engines, and custom instructions into real speedups on real programs and ML models. We need a toolchain that keeps pace with an AI-driven hardware design loop, one that can test proposed architecture features against real code within days and feed what it learns back into the chip.
You will own the compiler toolchain for our CPU and accelerator cores end to end: backend code generation, support for new instructions, lowering ML models to hardware, and performance. As a Staff engineer you will set technical direction, make the key architectural calls, and raise the engineering bar for the team. This is a small and early team, so the methodology and infrastructure are still being defined.
WHAT YOU WILL DO
- Own the design and implementation of the compiler backend for our CPU and accelerator cores, including instruction selection, scheduling, register allocation, and target-specific optimization.
- Add support for standard and custom instruction set extensions (vector, matrix, bit-manipulation, and domain-specific) across the compiler, assembler, linker, intrinsics, and binutils.
- Build the path from ML frameworks (PyTorch, JAX) through MLIR to our hardware, covering dialects, tiling, data layout, vectorization, and kernel generation.
- Lead hardware/software co-design: evaluate proposed ISA and microarchitecture features by compiling real workloads against simulators and performance models, and report quantitative results to the architecture team.
- Own performance and correctness infrastructure: benchmark suites, differential testing, fuzzing, and CI against simulators, FPGA, and silicon.
- Apply ML to the compiler itself through learned cost models, autotuning, and ML-guided optimization heuristics.
- Set the technical roadmap for the compiler effort, mentor engineers, and upstream work to the open-source projects we build on.
MINIMUM QUALIFICATIONS
- BS in Computer Science, Computer Engineering, Electrical Engineering, or a closely related technical field.
- 8+ years of compiler development, including substantial hands-on backend and code generation work for a CPU or accelerator target.
- Expert C++ and deep knowledge of LLVM (or GCC) internals.
- A track record of shipping a production compiler toolchain for real hardware: new target bring-up, support for new ISA extensions, or optimization work in a shipped compiler.
- Experience leading multi-quarter technical efforts across several engineers or teams.
PREFERRED QUALIFICATIONS
- MS or PhD with a focus on compilers, programming languages, or computer architecture.
- Experience with vector/SIMD ISAs, ML accelerators (TPUs, tensor cores, NPUs), or DSPs, and with open-standard ISA ecosystems.
- Hands-on experience with MLIR, XLA, TVM, Triton, IREE, or torch.compile, and with kernel performance tuning.
- Experience applying ML to compilers (e.g. MLGO, learned cost models, autotuning) or building ML systems.
- Publications at PLDI, CGO, CC, ASPLOS, MICRO, ISCA, or MLSys.
- Upstream contributions to, or code ownership in, LLVM, MLIR, GCC, or binutils.
- Experience with architecture simulators (e.g. gem5) and performance modeling.