Shridipa Dhar

I’m a Machine Learning Lead and Research Intern focused on neural network optimization, LLM/RAG pipelines, and measurable performance improvements. Currently, I’m working at the Indian Statistical Institute (ISI), Kolkata, on Neural Networks.

In my ISI research, I engineered high-throughput PyTorch benchmarking pipelines, boosting experimental reproducibility by 40% across heterogeneous deep learning architectures, and optimized architecture evaluation speed by 25% using automated, multi-variable profiling scripts to diagnose CNN and vision transformer bottlenecks. At KIIT Nexus Society, I standardized end-to-end MLOps tracking workflows for student research cells, accelerating deployment velocity by 30%, and led a structured curriculum on LLM fine-tuning and RAG optimization for 80+ engineering students.

My project work centers on building systems that balance accuracy, latency, and reliability: Research Mind raised semantic retrieval accuracy by 28% with a modular RAG pipeline, while cutting ingestion latency by 35% using concurrent asynchronous PDF parsing and indexing. I’ve also reduced LLM domain hallucinations by 18% in Medic AI using medically-structured graph knowledge bases, and improved adaptive learning precision by 22% with a performance-based behavioral model. I bring a strong technical writing mindset from publishing on Hashnode and a solid CS foundation through CS50’s AI and CS50x.