Expert Answer
Anonymous
As a EM who has successfully delivered multiple complex machine learning and AI systems leading, coordinating and influencing cross functional orgs and teams of engineers, scientists and TPMs, I have always been at forefront of anticipating problems while setting the direction, and managing and executing the plan. For example, in one of the current projects that i am leading, the project is to develop a conversational generative AI voice assistant to support digital and device customers of Amazon, i called out proactively some of the critical issues in the design proposal on how multi-turn conversations will be orchestrated between LLM based domain specific reasoning engine and general purpose speech to speech engine which is capable of speech recognition, chit-chat and speech rendering. The issue called out was mainly around the arbitration design between the two engines for different types of conversation including chit-chat and the ones that would involve a high degree of reasoning. Upon further discussion and brainstorming, we settled down with a design where reasoning engine would assume the full responsibility for all types of conversations that would ensure a consistent CX. the new design also simplified out latency mitigation efforts and integration challenges.
