Milin Bhade
Compilers @AMD | M.Tech CSA, IISc Bangalore
Bengaluru
3+ years of work experience
Jan 20268 mosPresent
Bengaluru, Karnataka, India
Aug 20241 yr 7 mosFeb 2026
Bengaluru, Karnataka, India
Worked on integrating OpenXLA with a custom Processing-Near-Memory (PNM) accelerator backend to enable efficient execution of ML workloads for memory bound workloads
Jul 20231 yr 1 moJul 2024
Bengaluru, Karnataka, India
Focused on optimizing transformer model compression through reinforcement learning-driven pruning, using models like BERT and LLaMA2. Conducted extensive research on pruning algorithms, designed and implemented custom solutions, and rigorously tested their performance across key benchmarks.
1. Compared various methods (RL-TriM, LLM-Pruner, and Uniform) with different pruning criteria (L1, L2, and Wanda), where RL-TriM Wanda achieved the best average benchmark score (0.5985) with 20% pruning and 0.26% drop in average score across LM-eval-harness benchmark
2. Automatic pruning of transformers based model for MLPs and Attention heads, with minimal performance impact using DDPG RL agent.
Milin’s AI Stack
Education
Master of Technology - MTech, Computer Science and Automation
Indian Institute of Science (IISc)
20222 yrs2024
Bachelor of Technology - BTech, Electrical and Electronics Engineering
Government College of Engineering, Amravati.
20174 yrs2021
Join Milin and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal Boss