Developer Screening Challenge (Mandatory) To ensure we are building this system at a high architectural and engineering level, all applicants must complete the following screening steps. Applications without these requirements will not be reviewed. --- Phase 1 – Architecture Design (Required) Submit a short technical document (maximum 3 pages) explaining: 1. How you would architect an AI system capable of: * Autonomous multi-platform publishing * Human-like comment and DM engagement * Conversation memory retention * Risk mitigation and unsafe response prevention 2. Include: * Basic architecture diagram * Proposed cloud infrastructure * API rate limit management strategy * Fail-safe and system recovery design We are specifically evaluating architectural thinking, not generic descriptions. --- Phase 2 – Mini Proof of Concept (5–7 Days) Build a small working prototype that: 1. Accepts a single text post input 2. Generates 3 stylistic variations: * Professional * Authority-driven * Casual 3. Generates 5 possible auto-replies to user comments 4. Includes a basic content moderation or risk filtering mechanism No live social media integration required. Code must be clean, modular, and documented (README included). --- Phase 3 – Risk Intelligence Scenario Provide a written explanation: If the system detects an aggressive, controversial, or legally sensitive comment — how does your architecture prevent unsafe automated responses? We expect discussion of: * Risk scoring * Confidence thresholds * Human approval fallback * Logging & audit systems * Moderation layers Simple rule-based filtering is not sufficient. --- ## Phase 4 – Adaptive Learning Strategy Explain how you would design the system to learn and refine brand voice over time. We are looking for approaches involving: * Embeddings * Vector databases * Retrieval-Augmented Generation (RAG) * Style modeling * Feedback loops * Memory persistence --- Important Candidates who do not submit: * An architecture explanation * A diagram * And a structured technical proposal Will not be considered.