Vivasoft AI Scene
It has become a prerequisite for companies to develop custom software products to stay competitive.
Here are the stuff we are working on right now
1. Orange Agent: AI Customer Service Solution
Objective
Develop a robust Retrieval-Augmented Generation (RAG)-based system for customer agents to streamline responses to queries using both structured and unstructured organizational knowledge base.
Key Achievements
- Designed and implemented a scalable architecture using RAG to process and respond to customer agent inquiries efficiently.
- Customized an intricate pipeline to handle unstructured data, ensuring accurate and contextually relevant responses.
- Deployment initiated for one of the largest telecom companies in Bangladesh, setting a precedent for AI-driven customer service in the region.
Impact
This project leverages organizational knowledge effectively, reducing manual intervention and improving customer satisfaction by delivering consistent and precise information as quickly as possible.
2. Ollivander: Product Recommendation Platform
Objective
Build a natural language query-based product recommendation platform to enhance user experience in e-commerce environments.
Key Achievements
- Developed a recommendation engine powered by natural language understanding to interpret user queries and suggest relevant products.
- Indexed product data in a vector database, enabling RAG to ensure high precision in responses.
- Enhanced search relevance, resulting in a seamless and intuitive shopping experience.
Impact
The Ollivander platform transforms product discovery, bridging the gap between user intent and available inventory through intelligent recommendations.
3. Previa: Conversational Interviewing Platform
Objective
Create an AI-powered platform capable of conducting interviews and providing feedback, aimed at streamlining recruitment processes in order to create an efficient pre-screening.
Key Achievements
- Designed an AI conversational agent to simulate human interviewers, ensuring objective and consistent assessments.
- Successfully conducted nearly 50 candidate interviews as part of the Weekdemy program selection process.
- Automated feedback generation to assist hiring managers in evaluating candidate performance efficiently.
Impact
Previa reduces recruitment time by providing an insightful representation of the candidate and while ensuring a standardized evaluation process, improving the overall quality of talent acquisition.
4. Auto Query Agent: Autonomous Query Management System
Objective
Develop a multi-agent AI system for generating, executing, and resolving massive database queries autonomously.
Key Achievements
- Leveraged RAG techniques for advanced context management, enabling the system to handle complex queries dynamically.
- Designed error-resolution mechanisms for self-correction, improving reliability and reducing manual intervention.
- Conducted extensive R&D to optimize system performance and enhance decision-making capabilities.
Impact
The Auto Query Agent sets the foundation for autonomous systems capable of managing and resolving intricate tasks, thereby streamlining operations and reducing costs.
5. Automated Speech Recognition with Bangla
Objective
Advance Bangla Automated Speech Recognition (ASR) systems to address challenges unique to the language, such as poor audio quality and diverse dialects.
Key Achievements
- Conducted extensive research using state-of-the-art models like Whisper and Wav2Vec2 for Bangla ASR.
- Focused on fine-tuning models and preparing datasets to improve recognition accuracy in challenging audio conditions.
- Proposed solutions to enhance model performance for low-resource languages.
Impact
This initiative contributes to the development of accessible ASR systems for Bangla, fostering inclusivity and bridging the digital divide in underrepresented linguistic communities.
6. Vin : Transforming interface development workflow
Objective
Develop a design and collaboration tool that enables users to intuitively build web applications with minimal effort, generating framework-agnostic codebases with the assistance of large language models (LLMs).
Key Achievements
- Designed and implemented a system that automatically analyzes and structures complex interface components, improving the efficiency of building web applications.
- Developed a feature that seamlessly structures interface units, ensuring flexibility, scalability, and adaptability.
- Leveraged LLMs to generate clean, framework-agnostic solutions, automating the creation of well-structured, maintainable codebases tailored to the user’s needs.
Impact
This project simplifies the application development workflow, allowing developers to quickly create dynamic applications with minimal intervention. The integration of LLMs ensures high-quality, adaptable code that can be seamlessly deployed across different platforms.
Accelerate Your Software Development Potential with Us
