Dr. Rishabh Upadhyay

Developed and improved a health information retrieval model using advanced NLP techniques like Natural Language Inference (NLI). Employed cosine similarity and an NLI-based stance detection model to compute a unique 'genuineness score' that assesses the credibility of health information, resulting in significant retrieval accuracy improvements. Worked extensively with Python, PyTorch, Transformers, BioBERT, and SciFIVE for NLI and biomedical text mining tasks.

Augmented an existing health information retrieval model, focusing on increasing its transparency through explainability features. Integrated a comprehensive database of medical journal articles, providing detailed explanations in the form of evidences for retrieved health information. Utilized Python, PyTorch, Transformers, and BioBERT for biomedical text mining tasks.

Developed and trained specialized Speech-To-Text (STT) models for ten low-resource languages, employing advanced techniques such as transfer learning and language-specific data augmentation. Created a scalable speech processing pipeline capable of handling over 1 billion audio recordings daily, leveraging cloud infrastructure and parallel processing techniques. Used Python (Pandas, Scikit-learn),, Git, and Kaldi for data analysis, machine learning, system operations, and speech recognition tasks.

Developed pcStream2, a variant of pcStream, employing windowing and persistence techniques along with a novel IPCA algorithm called Just-In-Time PCA (JIT-PCA) for data stream clustering. Contributed to a research paper published in the 32nd ACM/SIGAPP Symposium On Applied Computing (SAC), showcasing project outcomes and advancements. Utilized R on the Ubuntu Linux platform for algorithm development, data manipulation, and evaluation.

Implemented an advanced algorithm to extract meaningful knowledge from academic documents employing Natural Language Processing techniques. The developed knowledge extraction algorithm resulted in publications in the Federated Conference on Computer Science and Information Systems (FedCSIS) and the Portland International Conference on Management of Engineering and Technology (PICMET)

- DoSSIER H2020 ITN funding (Domain Specific Systems for Information Extraction and Retrieval). - Will be working on Assessing Credibility, Value, and Relevance from health related content.

This program consists of theoretical (1 year) and practical knowledge (1 year). List of Recent Courses: - Finite Optimisation. - Advance Machine Learning. - Randomised Algorithm.

Machine learning Data Mining Intelligence system