Worked on the pipecat-ai/pipecat repository to deliver the InceptionLLMService integration with Mercury-2, enhancing conversational AI capabilities through configurable reasoning effort and real-time processing. Focused on backend development using Python, the work established an architecture that supports diffusion-based reasoning models and enables future scalability. By introducing adjustable latency and quality tradeoffs, the integration allows product teams to tune performance for different user needs. The API surface was aligned with existing services to facilitate maintainability and deployment. No major bugs were addressed during this period, with efforts concentrated on feature delivery and optimizing response times for Mercury-2 powered conversations.