Ojasv implemented BERT-based sequential sentence classification and LDA topic modelling pipelines for my NLP research at Radboud. The tokenisation logic, attention-mask handling, and hyperparameter sweep were all production-quality. Invaluable for my PhD thesis.

Ojasv delivered a full Boston Housing regression pipeline in R — feature engineering, multicollinearity diagnostics via VIF, stepwise AIC model selection, and residual analysis. He explained the OLS assumptions clearly enough that I could defend the methodology in my exam.

Ojasv handled cross-lingual scraping, bilingual Arabic-English preprocessing with custom tokenisation, and fine-tuned XLM-RoBERTa for multilingual sentiment classification. The F1 scores on Arabic test data exceeded our in-house baseline by a significant margin.

He walked me through logistic regression, decision trees, and ensemble methods in R for Business Analysis at SMU — tuning regularisation parameters, interpreting ROC-AUC curves, and writing reproducible R Markdown reports. Thorough, precise, and always on time.

Ojasv built custom Power BI DAX measures, parameterised SQL queries, and Python automation scripts tailored to State Street's analytical workflows — then walked me through each piece clearly enough to answer technical interview questions with confidence.

Ojasv independently delivered a deep learning pipeline for Stock Price Prediction — LSTM architecture with sliding-window sequence encoding, dropout regularisation, and backtesting — plus a full Real-Time Traffic Signal Optimisation thesis from scratch. Both were exceptional.

Ojasv implemented BERT-based sequential sentence classification and LDA topic modelling pipelines for my NLP research at Radboud. The tokenisation logic, attention-mask handling, and hyperparameter sweep were all production-quality. Invaluable for my PhD thesis.

Ojasv delivered a full Boston Housing regression pipeline in R — feature engineering, multicollinearity diagnostics via VIF, stepwise AIC model selection, and residual analysis. He explained the OLS assumptions clearly enough that I could defend the methodology in my exam.

Ojasv handled cross-lingual scraping, bilingual Arabic-English preprocessing with custom tokenisation, and fine-tuned XLM-RoBERTa for multilingual sentiment classification. The F1 scores on Arabic test data exceeded our in-house baseline by a significant margin.