Chatbot on the Edge

Supervisor: Sareh Maleki, Haleh Dizaji
Author: Luis Palma, Sulaiman Hazaa

Abstract

The Edge-AI project provides techniques for deploying intelligent systems on resource-constrained devices. Key research activities include model compression methods such as pruning and quantization, efficient embedding techniques for lightweight and effective data representation and retrieval, and secure, robust architectures for privacy-preserving, on-device inference. The scope of application spans various domains, including environmental monitoring, smart sensing systems, interactive interfaces, and compact conversational agents.

One of the tasks of this project involves the design and deployment of domain-specific question-answering systems—lightweight chatbots trained on curated domain-specific datasets and optimized for deployment on edge hardware. The internship offers practical experience in the end-to-end development pipeline, including data preparation and embedding, training compact language models, implementing retrieval-augmented generation (RAG) pipelines, and applying optimization strategies tailored for edge deployment. Interns will gain exposure to real-world system integration, applied machine learning workflows, and collaborative scientific research.