Performance Optimization & Feature Development for CRM Middleware (Node.js/TypeScript/Claude AI)

About Us We're a B2B SaaS company with a CRM middleware platform that syncs customer data between Salesforce, HubSpot, and our clients' internal systems. The platform handles tens of thousands of sync events per day across 40+ accounts. The entire system was built with Claude and uses several MCP (Model Context Protocol) servers to power AI features across the pipeline. The Situation Our middleware layer has hit a performance bottleneck. As account volume and data complexity have scaled, the sync pipeline can't keep up — jobs queue up, throughput degrades during peak hours, and large accounts starve resources for other tenants. We need a senior backend engineer with deep AI integration experience to identify and resolve the bottlenecks and deliver new functionality that brings the pipeline up to scale. Scope of Work - Diagnose and resolve pipeline throughput bottlenecks — likely refactoring from sequential to parallelized job execution with priority-based queuing - Profile and optimize slow PostgreSQL queries and indexing for accounts with 50k+ contacts where sync jobs currently time out - Implement field-level mapping configuration so clients can define custom sync rules and data transformations per CRM field - Build a sync history and audit log API (REST) with filtering and pagination - Extend our multi-tenant authorization layer to support API key scoping per integration (currently scoped at the account level) - Implement a background job for sync conflict detection and resolution (e.g., concurrent updates to the same record across systems) - Review and harden our existing MCP server integrations — we use MCP tools for data enrichment, entity resolution, and intelligent field mapping, and they need tuning as data volume grows - Ensure all Claude-powered features (structured outputs, tool use, prompt chains) are resilient under increased load and handle edge cases gracefully Tech Stack - Backend: Node.js, TypeScript, Express, with some NestJS services - API: REST + GraphQL (Apollo Server) - AI: Claude API, MCP servers (data enrichment, entity resolution, field mapping), structured outputs, tool use, function calling - Database: PostgreSQL (RDS), Redis for caching and rate limiting - Queue/Events: AWS SQS, Lambda functions for async processing - Infra: AWS (Lambda, API Gateway, S3, RDS, SQS, CloudWatch) - CI/CD: GitHub Actions, Docker - Testing: Jest, some Playwright for integration tests - Repo: GitHub, PR-based workflow with code review Requirements - Deep experience with Claude API, MCP, tool use, and AI integration in production systems — this platform was built on Claude and you need to be fluent in it - Strong Node.js and TypeScript — this is hands-on backend engineering, not prompt playground work - Solid PostgreSQL skills — query plans, indexing strategy, migrations, and performance tuning - Experience with message queues, background jobs, and async processing on AWS - Familiar with GraphQL APIs (Apollo Server preferred) - Ability to onboard quickly into an unfamiliar codebase and deliver with minimal ramp-up - Experience with multi-tenant SaaS architecture and authorization patterns - Writes tests. Not negotiable. - Clear async communication. We're a small distributed team and rely on written updates. Nice to Have - Experience with CRM integrations (Salesforce, HubSpot APIs) - Healthcare or enterprise SaaS background (some of our clients are in regulated industries) - Experience with LangChain, LangGraph, or agentic workflows - Familiarity with Redis caching strategies and cache invalidation patterns To Apply Tell us about a production system you've built or maintained that uses Claude or LLM integrations. What was the architecture, what broke at scale, and how did you resolve it? Skip the generic cover letter — we want to see how you think about AI in real systems.