Page 64 - tech fest 2025 אשדוד חוברת תקצירים

64
An Educational Chatbot Based on Generative AI: An Applied Research Project
IEM-A-03
Joseph Bush; yosefbu@ac.sce.ac.il
Advisor: Ms. Shira Schweitzer
SCE - Shamoon College of Engineering, Ashdod
This applied research project presents a prototype for an advanced educational chatbot powered by generative artificial intelligence. Designed to support learning in text-rich subjects, especially history, the chatbot operates fully offline (On-Device), without cloud or Internet dependency, making it suitable for secure environments, such as defense, education, and government sectors. The system is based on Meta’s open-source LLaMA 2 model and uses technologies like LoRA and BitsAndBytes for efficient fine-tuning and memory optimization. Due to budget constraints, training was done on Google Colab, while the system was developed for local deployment. The project combines practical development with a solid understanding of transformer architecture and attention mechanisms, resulting in a scalable, secure AI learning tool.
Keywords: applied research, attention mechanism, bitsandbytes, data privacy, educational chatbot, fine-tuning, generative AI, Google Colab, history education, information security, large language model, llama 2, LoRa, on-device, quantization, transformers
An Information System for Matching Caregivers and Patients in Home Health Care
IEM-A-04
Gaya Nessa Stam; gayalibby19@gmail.com Sapir Cohen; sapircn7@gmail.com
Advisor: Dr. Hagai Ilani
SCE - Shamoon College of Engineering, Ashdod
Elderly people eligible for long-term care benefits need daily help, close supervision, and monitoring due to a medical and functional condition that puts them or others at risk. Therefore, they must find suitable caregivers. Currently, this process involves bureaucracy, is time-consuming, and requires efforts made by the elderly and their families. Even after finding a caregiver, the match is not always ideal, leading to frustration and frequent replacements.
We have designed an interface for a smart system that matches caregivers and patients based on an advanced matching algorithm. The algorithm considers various parameters to find the optimal match for each caregiver and patient.
Keywords: assignment problems, home medical care, Hungarian algorithm, stable matching