Soubhagya Saxena

I built and compared ANN and CNN models for six-class scene classification using PyTorch, then benchmarked their accuracy, training time, and parameter counts.

For sentiment analysis, I trained and compared RNN, LSTM, and GRU models on clothing reviews. The GRU achieved 92.2% test accuracy.

My projects also include time-series outlier detection using rolling statistics and a 2-sigma rule, plus a student data analyzer with cosine-similarity matching.

At Uplyx Solutions, I completed structured AI training and worked on live, industry-oriented projects applying machine learning and data analysis. I also gained exposure to backend fundamentals, IT operations, system documentation, and data management as a Backend Support – IT Intern at TDS Management Consultant Pvt Ltd.