Manu Nicholas Jacob

Bio:

Manu Nicholas Jacob is an independent researcher who measures what governs AI inference performance and energy, from 2 GB Raspberry Pi 5 boards to x86 servers and cloud GPUs. He maintains two open source tools on PyPI: llama-roofline, which tells llama.cpp users whether decode is memory-bandwidth-bound, and ml-systems-lab, a benchmark framework spanning SSH, edge and Kubernetes devices. He contributes to AIPerf, the NVIDIA Dynamo benchmarking tool, served on the USENIX ATC '26 Artifact Evaluation Committee, reviews for the Journal of Open Source Software, and is an accepted speaker at the 2027 Embedded Vision Summit. He holds a dual B.S. in Electrical and Computer Engineering from UMass Amherst.