Senior Backend & AI Agent Architect (Next.js, TypeScript, Python) — Long-Term Project

Job Description: (No agencies please) We are looking for a heavy-hitting Senior Backend Architect (8+ years of production experience) with a deep background in designing unified API integration layers and writing tools for agentic AI architectures. You will initially lead a rapid intensive sprint to implement an Ayrshare integration within Juliet—an AI-driven website and growth builder tool. The immediate goal is to wire up our baseline "Ship v1" social publishing spine, enabling profile connecting, scheduling, publishing, and webhook-driven monitoring across multiple brands. Following a successful kickoff, this will transition into a long-term engagement where you will lead the architecture of our broader, manifest-driven unified integration layer (building swappable provider adapters for CRM, SEO data, and messaging infrastructure). The initial codebase is already partially prepared: the database schemas, state lifecycles (draft → scheduled → published), and front-end tabs exist. Your job is to hit the ground running, act as the primary architect to build a clean, swappable PublishAdapter layer, write robust webhook ingestion engines, and cleanly expose backend capabilities as tools into our LangGraph-based AI agent layer. Key Responsibilities: Architect Swappable Integration Layers: Design and implement modular, provider-abstracted integration adapters (TypeScript/Node.js) behind strict boundaries, ensuring we avoid hard-coupled vendor lock-in. Build Agentic AI Tools: Cleanly translate complex backend API logic, schema validation, and multi-tenant connection states into autonomous, executable tools (Python/LangGraph) optimized for LLMs. Bulletproof Inbound/Outbound Data Pipelines: Implement HMAC-verified webhook ingestion, deep idempotency strategies (idempotencyKey), and rate-limit/backoff handling to eliminate data duplication or silent connection failures. Accelerated Development: Work comfortably alongside agentic developer tools (such as Claude Code and MCP workflows) to rapidly prototype, validate payloads, and push production code within tight constraints. Technical Requirements: Seniority: 8+ years of professional engineering experience architecting multi-tenant B2B SaaS platforms or complex data-integration engines. Languages: Dual-fluency in both TypeScript (Next.js/Node.js) for the backend/BFF API routes and Python for our core AI workflow layer. AI Agent Context: Practical understanding of stateful multi-agent frameworks (e.g., LangGraph, LangChain), human-in-the-loop interrupts, and dynamic tool schemas. API & Security Master: Deep expertise handling complex OAuth loops, securely managing encrypted tenant tokens (AES-256-GCM), and handling tricky platform-specific formatting quirks (carousels, media file uploads, custom CTAs). Integration & Tooling Context (Ayrshare Primitives): To help accelerate this fast-turnaround sprint, please review the following architectural primitives from the Ayrshare ecosystem: Historical post data and analytics are fully accessible via the API, allowing you to run that data through our internal modeling for tone-of-voice or performance analysis. You can check the documentation details for the History API [https://www.ayrshare.com/docs/apis/history/overview] and the Analytics API [https://www.ayrshare.com/docs/apis/analytics/overview]. Because our agent infrastructure runs on LangChain and Claude, you can leverage Ayrshare's live MCP Server [https://www.ayrshare.com/docs/additional/mcp-action-server] and their specialized Claude plugin [https://www.ayrshare.com/docs/additional/mcp-claude-code-plugin] as developer-time accelerators. We will be hooking into a data repository utilizing webhooks to push data back to our platform and reduce constant API polling. The receiver architecture can be framed around their Webhooks API [https://www.ayrshare.com/docs/apis/webhooks/overview]. All infrastructure runs on an enterprise-grade 99.99% SLA with dedicated engineering resources. All tiers include full feature access, with the exception of the vendor's Max Pack and FB Boosted ads features. Project Context: First Sprint: 1–2 days to build the baseline social publishing spine within the Juliet builder platform. Long-Term Horizon: Designing a manifest-driven, generic router to rapidly expand our API catalog across analytics, CRM routing, and complex external data endpoints. Starting Line: Detailed PRDs, database schema blueprints, and pre-mapped endpoint matrices are ready for a clean Day 0 start. If you are a high-autonomy backend engineer who builds production systems that simply don't break, and you're excited to rapidly architect the bridge between external APIs and autonomous AI agents, apply with a brief overview of your technical background. *****Required Screening Questions***** 1. Ayrshare requires a public URL to publish media, but its own hosted file URLs expire after 30 days. If a user schedules a post 45 days out using a Juliet Studio asset, how would you configure our storage lifecycle and the PublishAdapter to ensure the post doesn't fail on day 45 without leaving permanent file bloat in our system? 2. Ayrshare's PATCH /post endpoint only allows you to update the scheduled time or approval state—any changes to the caption or media require a full delete-and-recreate. How would you handle a user's content edit in our database queue so that we seamlessly transition the platform receipts to the new Ayrshare ID without creating ghost/duplicate entries on the user's calendar?