MohammadSajad Alipour
About
I am a PhD student in Computer Science at Rensselaer Polytechnic Institute. My research focuses on building efficient and scalable learning systems, with an emphasis on model compression techniques such as pruning, low-rank approximation, and quantization, as well as low-rank adaptation, model merging, data valuation and coreset selection. I earned my Bachelor’s degree in Computer Engineering from Amirkabir University of Technology.
Research Interests
- Model compression
- Model merging
- Efficient inference for large language models
- Efficient fine-tuning and alignment
- Data valuation
- Coreset selection
Publications
Open-source Projects
Implemented a Persian grapheme-to-phoneme model with a sequence-to-sequence transformer. Trained two variants and evaluated on a test set. Code
Used IPOPT and Pyomo to search for 2×2 matrix multiplication algorithms with 7 scalar multiplications (inspired by AlphaTensor). Code
Built an inverted-index search engine with TF-IDF ranking and optimization techniques, and extended it with Elasticsearch. Code