Anisha Shende

Anisha Shende

she/her

Undergrad at VIIT

Posts by Collection

projects

publications

Building and Optimizing a LLaMA-Based Chatbot for Academic Support

Published in Journal 1, 2024

This study explores a chatbot developed using the LLaMA framework, specifically tailored to address academic queries at VIIT. To generate contextually appropriate responses, the system integrates information from MongoDB using a combination of embedding techniques and tools from Hugging Face. By leveraging MongoDB, the chatbot retrieves relevant institute-related information and produces clear and coherent answers through carefully fine-tuned LLaMA models. To evaluate the effectiveness of different models in delivering accurate and pertinent responses, a comparative analysis is conducted. This chatbot aims to assist students in real-time, enhancing communication with the institution and facilitating a smoother, more efficient user experience.

Recommended citation: Aditya Wanve, Siddhesh Raskar, Anisha Shende, Aditya Patil, Anuratan Bahadure, Manisha Mali, 'Building and Optimizing a LLaMA-Based Chatbot for Academic Support,' in International Journal of Research in Engineering, Science and Management, vol. 7, no. 11, pp. 52-56, November 2024.
Download Paper

Predicting Optimal Joint Angles for Prosthetic Knee Design Using Gait Analysis

Published in Modern Digital Approaches to Care Technologies for Individuals with Disabilities, 2025

Gait analysis significantly enhances the quality of life for limb amputees. Conventional techniques require a considerable amount of effort and time. This study uses machine learning and gait data to identify the optimal knee joint angles for prostheses. A total of 34 individuals, including 14 unilateral above-knee amputees, had their joint angles and other characteristics analyzed using models such as Bidirectional LSTM, ANN, RNN, Decision Tree, Random Forest, Gradient Boosting, and SVM. The Random Forest model outperformed other methods, achieving R² values of 0.95 for testing and 0.96 for validation. The findings highlight how machine learning can drive advancements in prosthetic knee customization for improved comfort and natural walking patterns. Future studies aim to refine these models and explore additional data sources to enhance prosthetic knee design accuracy.

Recommended citation: Shende, A. A., Rathi, S. R., & Mahalle, P. N. (2025). Predicting optimal joint angles for prosthetic knee design using gait analysis. In Modern Digital Approaches to Care Technologies for Individuals with Disabilities. IGI Global.
Download Paper | Download Slides