AI AGENT solution maneger

Job Title: Senior Full-Stack Developer for AI Companion App (Focus on Streaming & Safety) Project Overview:
We are building an advanced AI companion application. We need an experienced developer to handle critical frontend-backend integration tasks, specifically focusing on real-time streaming stability and implementing a robust safety guardrail system. Key Responsibilities & Technical Requirements: 1. Data Structure Alignment
You must ensure the frontend strictly adapts to the backend response structure. We require the implementation of a frontend adapter layer to isolate raw backend data formats. This ensures that if backend fields change, the UI logic remains stable and does not break. 2. Robust Streaming Communication (SSE/WebSocket)
This is a high-priority task. You must implement an engineering-level communication wrapper that includes: ● End-of-Stream Indicators: The backend must push a specific signal to tell the frontend when the stream ends, preventing the UI from hanging indefinitely. ● Stability Mechanisms: Implement timeout detection (e.g., 30 seconds), smart retry mechanisms, heartbeat keep-alives (pinging every 15 seconds), and auto-reconnection features to handle network fluctuations smoothly. 3. Unified Exception Handling
Do not rely on scattered error checks. You need to integrate error handling into a unified request layer (like a safeRequest utility). Both success and failure responses must return a standardized structure so the UI can gracefully handle empty states or timeouts without freezing. 4. Risk Control and Safety Guardrails Integration
This is crucial for our AI product. You will be responsible for integrating safety filters that prevent harmful content while maintaining immersion. ● Standardized Blocked Payloads: When the backend blocks a message, it must return a specific JSON format indicating the block reason and a fallback message. The frontend must render this fallback message instead of crashing. ● Dual-Layer Validation: The architecture must support both input filters (checking user prompts) and output evaluators (scanning AI responses) before they reach the user. ● Crisis Intervention: If the system detects self-harm or extreme distress, the frontend must immediately intercept the standard chat flow and display a crisis support interface with hotline numbers. 5. Context Management
You must implement session state isolation to prevent the AI from hallucinating or reusing old records incorrectly. The system should allow forcing a new session to clear historical interference when necessary. What We Are Looking For: ● Proven experience with LLM integration and streaming protocols (SSE/WebSocket). ● Strong understanding of AI safety patterns and content moderation APIs. ● Ability to write clean, defensive code that handles edge cases and network errors gracefully.