Big Mango provides AI assistants that act as force multipliers for their users. Oivan partnered with Big Mango to mature their engineering ecosystem, creating a comprehensive roadmap to transition their “stealth-mode” capabilities into a reliable, enterprise-grade production environment.
Big Mango operates at the bleeding edge of the productivity sector, providing AI assistants that act as force multipliers for their users. As the demand for generative AI grew, Big Mango needed to ensure their underlying architecture was not only robust but capable of evolving alongside rapidly changing LLM technologies. We partnered with Big Mango to architect their engineering ecosystem, creating a comprehensive roadmap to transition their “stealth-mode” capabilities into a reliable, enterprise-grade production environment.
Big Mango is in the business of redefining personal and professional productivity. By leveraging advanced Artificial Intelligence, they provide cutting-edge assistants designed to seamlessly integrate into daily workflows.
However, in the volatile world of AI development, reliability is paramount. Big Mango recognized that to maintain their market lead, they needed to transition their codebase from a feature-focused prototype to a test-driven, scalable platform capable of supporting AI-augmented development processes.
The collaboration began with a focus on stability and scalability. To align the project with high-velocity AI development, Oivan took charge of re-architecting the core system.
The initial priority was refactoring the code base to achieve high test coverage. This step was crucial to create a safe environment for AI-augmented development, ensuring that the code was not just functional, but clean enough for AI agents to assist in future maintenance and expansion without introducing regressions.
To solve the challenge of evaluating subjective AI responses and rapidly evolving models, we engineered a custom “LLM Testbed.”This infrastructure acts as the engineering backbone, allowing Big Mango to capture ideal customer behaviors and rigorously test them against various configurations. The team can now swap out and mix distinct LLM models, voice models, RAG approaches, and VOIP settings with ease. The impact on agility was immediate: the new architecture reduced model swapping time from days to minutes.
To ensure long-term fidelity, we integrated the IBM WatsonX Evaluation platform. This allowed us to create baseline Query/Response metrics, positioning the project for automated regression testing.
The project highlighted the necessity of specialized infrastructure in AI development. The LLM Testbed allowed for real-time and side-by-side comparison of impact on responsiveness and fidelity, giving the stakeholders immediate visual and data-driven feedback on how model changes affected the end-user experience.
By combining the flexibility of the Testbed with the rigorous metrics of WatsonX, we moved beyond subjective testing to a continuous monitoring approach, ensuring that Big Mango’s production quality remains consistent even as they scale.
Q1: How does Oivan help AI companies scale from prototype to production?
Oivan re-architects AI systems for enterprise-grade reliability. We have refactored numerous codebases with high test coverage, creating a stable foundation for AI-augmented development that evolves alongside rapidly changing LLM technologies.
Q2: What is Oivan LLM Testbed and how does it accelerate AI development?
Oivan LLM Testbed is custom infrastructure for evaluating AI model performance. Oivan’s solution enables rapid testing of different LLM models and RAG approaches—reducing model swapping time from days to minutes.
Q3: How can Oivan support my company?
Oivan is a global consultancy, offering end-to-end AI development support including AI/ML implementation, cloud infrastructure (AWS, Azure, Google Cloud, STC Cloud), cybersecurity, and DevOps services.
