Summary
Parameswaran Raman is a Research Scientist with 13 years of experience designing and optimizing large-scale machine learning systems, currently focused on algorithms for pre-training LLMs at Meta. He has led efficiency and systems co-design work for AWS Deep-Engine Science at Amazon, developing large-batch training, 3D parallelism and optimization techniques that accelerate real-world LLM pre-training. His academic work includes scalable, distributed inference and optimization methods (DS-FACTO, ESVI, DS-MLR) from a PhD in Machine Learning, giving him deep expertise in hybrid model-data parallelism and lock-free asynchronous algorithms. He blends research rigor with production impact—shipping dialog and recommender models for Alexa and Twitch earlier in his career—and has a track record of turning distributed ML theory into deployable systems. Based in Menlo Park, he brings rare fluency across statistical machine learning, systems engineering and practical large-scale training tricks that materially cut cost and time to train.
13 years of coding experience
16 years of employment as a software developer
Masters Computer Science, Masters Computer Science at Georgia Institute of Technology
Doctor of Philosophy (PhD) Machine Learning, Doctor of Philosophy (PhD) Machine Learning at Purdue University
University of California Santa Cruz
Masters (Integrated) Software Engineering, Masters (Integrated) Software Engineering at PSG College of Technology
English, Hindi, Tamil