We are seeking a senior-level Python developer to take over the development and maintenance of a sophisticated, high-context AI conversational agent. The project features a unique architecture combining Telegram’s User API with advanced LLM memory management and multi-modal capabilities. The Stack: - Core: Python (Asyncio), FastAPI. - Telegram: Telethon / Pyrogram (Expert level in User Accounts / MTProto is a MUST). - AI: OpenAI API (Assistants v2), OpenRouter, ElevenLabs (Voice), Vision. - Data: Airtable API (Primary DB), DynamoDB, Redis (for buffering). - Infrastructure: AWS (EC2/Lambda/S3), Docker, Ubuntu, Hatchet workflows. Key Responsibilities: - Token Optimization: Implementing Context Caching strategies (Static Prefixing) to manage high-token payloads (10k+). - State Management: Maintaining a complex memory-retention system (Memory-Lite) across multiple databases. - Media Handling: Managing asynchronous voice-to-text (Whisper) and image analysis workflows. - Persona Consistency: Ensuring the agent adheres to complex behavioral guidelines (High-EQ) without "breaking character." Critical Requirements (Red Flags): - DO NOT APPLY if you have only worked with Telegram BotFather accounts. We use User Accounts (MTProto). - DO NOT APPLY if you lack experience with Airtable as a primary backend for AI agents. - Experience with Prompt Layering and managing 1M+ token context windows is highly preferred. - Familiarity with AWS deployment and monitoring of Python services. How to Apply: Please provide a brief summary of your experience with Telethon (User Accounts) and Context Caching. Thank you!