toronto, on · --:--
hi! i study computer engineering at the
university of waterloo.
i like making ai models run fast on gpus.
currently
previously
-
↳
ai software developer @
ramsoft
ported and deployed ml models with tensorrt on triton inference server.
projects
-
↳
memory-bound llm inference
hand-written int8 cuda kernels that beat cublas on skinny shapes, plus a llama.cpp issue that sped up prefill 1.5×.
- ↳
- ↳