ConverseAI: Revolutionizing Edge AI and Vision with customizable models
Mohit Manjalkar
Department of Computer Engineering
Atharva College of Engineering
Mumbai, India
manjalkarmohit-cpmn@atharva.coe.ac.in
Sahil Bhojekar
Department of Computer Engineering
Atharva College of Engineering
Mumbai, India
bhojekarsahil-cpmn@atharva.coe.ac.in
Sonal Baury
Department of Computer Engineering
Atharva College of Engineering
Mumbai, India
baurysonal-cmpn@atharvacoe.ac.in
Dhanush Chandran
Department of Computer Engineering
Atharva College of Engineering
Mumbai, India
chandrandhanush-cmpn@atharvacoe.ac.in
Bhavna Arora
Department of Computer Engineering
Atharva College of Engineering
Mumbai, India
arorabhavna-cpmn@atharva.coe.ac.in
Abstract— ConverseAI is a robust, locally deployed artificial intelligence platform specifically designed to provide effortless interaction with optimized Large Language Models (LLMs), and its base engine is Ollama. With a focus on user privacy, personalization, and operation efficiency, it enables users to perform different tasks without the need for continuous internet connectivity. The web-based interface enables multimodal interaction in the form of text typing, vision-based inputs to handle images, and speech-to-text functionality for hands-free interaction. With Retrieval-Augmented Generation (RAG) built into it, ConverseAI enhances response precision with the use of document retrieval and embeddings to make responses relevant to the context. Its scalable design enables the deployment of various sizes of models and domain-specific variations, hence supporting a wide variety of workflows, which vary from educational support and professional automation to creating creative content. In successfully bridging the gap between artificial intelligence capabilities and real-world utilization, ConverseAI offers a very versatile and efficient AI experience.
Keywords— large language models, customization, efficiency, offline ai, web-based interface, multimodal interactions, text prompts, image analysis, speech-to-text, hands-free usability, retrieval-augmented generation, document retrieval
