Mikhail Lesin

Designed dialogue logic for voice AI assistants across different business scenarios, including multi-step conversational flows. Built and optimized system prompts for LLM behavior and improved conversation stability by analyzing real dialogue outcomes. Implemented integrations with amoCRM and Bitrix24 and worked with REST API/webhook style interfaces to connect conversational services with external systems. Coordinated with developers and clients throughout implementation and testing.

Independently developed a voice AI agent on the Vapi platform with an end-to-end dialogue architecture and system prompt. Selected an LLM/TTS combination through comparative testing and added protection against prompt injection. Implemented appointment scheduling, contact data collection, and lead qualification, then iteratively improved behavior based on test calls.

Built a conversational sales assistant focused on lead qualification and multi-stage sales conversations. Implemented handling of common objections and routing of qualified leads to sales managers. Tested different dialogue scenarios to tune model responses for smoother lead progression.

Designed the conversational architecture and a system prompt for a multi-branch consultation flow in a dental setting. Implemented handling for clarifying questions and structured data collection for clinic administrators. Tested multiple user scenarios and iterated on model behavior to improve reliability of the dialogue.