PhD Candidate · Imperial College London
Nicholas Fry
I work on differentiable rendering, geometric 3D vision, on-sensor vision, computer architecture, and compilers. I am supervised by Paul H. J. Kelly and Andrew J. Davison.
I work on differentiable rendering, geometric 3D vision, on-sensor vision, computer architecture, and compilers. I am supervised by Paul H. J. Kelly and Andrew J. Davison.
On-sensor computer vision, 3D vision, architecture, compilers.
Built an MLIR-based tool that auto-generates instruction-set-simulator state transition functions from an accelerator design (Verilog/Chisel → CIRCT → MLIR). Presented as a poster at EuroLLVM 2024.
Scheduling model for the IPU backend in TableGen/C++, and an MIR pass that removed register hazard bubbles. Reduced cycle count by up to 30% on client code. llvm-mca support and objdump work.
Visual Computing & Robotics, first-class honours. Thesis: 3D Gaussian Splatting on a Graph Processor. Supervised by Andrew J. Davison, awarded the Corporate Partnership Prize for Technical Innovation.
TA and marker for Imperial Robotics and Advanced Computer Architecture courses. Poster presenter at EuroLLVM 2024 and REACH 2024. Volunteered for EGSR 2024. Attended ICVSS 2025. Also attended PAISS 2025 and ACACES 2024 (both grant-awarded).
English is my mother tongue but I speak fluent French and have working knowledge of Spanish and Portuguese.