Engineer with a soft spot for performance, correctness, and clean experiments. I work across ML and systems: from training/serving setups to the lower-level pieces that make them work.
- Making training & inference runs easier to reproduce (env scripts, small profilers, benchmarks)
- Distributed training & serving ergonomics
- Compilers/kernels when performance actually matters
I’m currently evaluating issues and planning first PRs in:
- PyTorch / JAX (distributed, compile/Inductor/XLA)
- vLLM / model-serving tools
- TVM / compiler tooling
- AWS Neuron samples/SDK (Trainium/Inferentia)
(I’ll link PRs here as they open/merge.)
- Starlink handover analysis — mobility + network performance. Link
- Respiratory sound classification — applied ML with a short paper. Link
- LinkedIn: Ritunjay Murali
- Email: ritunjaymurali@gmail.com