Chellappan

my roadmap for getting into ML research as a student in 3 months: month 1: foundations + implementation learn linear algebra, probability, calculus, and optimization basics get comfortable with PyTorch, autograd, and training loops implement an MLP, CNN, and basic attention from scratch reproduce a classic paper like ResNet on a smaller dataset document results, training curves, and what went wrong month 2: reading + reproducing research pick one area: LLMs, computer vision, efficient training, etc. read 5-8 papers to understand the research direction choose one recent paper with open-source code and accessible compute reproduce its main experiment and compare against reported results understand the implementation, baselines, and limitations month 3: experiments + research outreach run 2-3 meaningful ablations on the reproduced paper test one small modification based on a clear hypothesis compare results across seeds and analyze failures write a short technical report with plots and findings publish code on GitHub and contact 10-15 relevant research labs what would you change?especially curious what actual ML researchers and PhD students think.