
Backend/DevOps Engineer for Multi-Agent Automation System (n8n, Supabase, Node.js, Vault/KMS)
Upwork
Remoto
•9 horas atrás
•Nenhuma candidatura
Sobre
I´m looking for an experienced backend/DevOps engineer to build a multi-agent system orchestrated by n8n (self-hosted), capable of processing documents, segmenting operational data, and generating automated reports. The system must persist results in PostgreSQL (Supabase) databases with logical isolation and data encryption (TLS, AES-256, Vault/KMS). - Technical Overview The project involves implementing a multi-agent architecture integrated with n8n, where each agent handles a specific domain (legal, operational, meteorological, reporting). All agents will operate as independent stateless services, communicating via signed JSON payloads. The n8n layer acts purely as the workflow orchestrator — it triggers, validates, and routes events, while dedicated async workers handle the heavy AI/OCR workloads. - Agents involved in the workflow Gatekeeper (Intent Validation and Routing) Requests the operator to explicitly choose the workflow type before processing. Validates the coherence between document type and content, tags metadata (client_id, org_id, workflow_id), signs the payload, and publishes the task to the proper queue. Generates audit logs and orchestrates approval flow. Legal Agent (OCR/NLP and Compliance Extraction) Processes legal documents (contracts, deeds, land registry files, etc.). Performs OCR (AWS Textract / GPT-4o Vision), extracts clauses, parties, areas, and dates, and cross-validates against CCIR, ITR, and CAR datasets. Returns a structured JSON payload to n8n, which persists it in the Legal Database (Supabase – schema-per-client) under AES-256 encryption and audit triggers. Operational Agent (Harvest, Management, and Daily Reports) Receives data from service orders, field sensors, spreadsheets, or ERP systems. Performs normalization and segmentation by operation type (planting, harvesting, spraying, maintenance) and time window (daily, weekly, monthly). Generates short daily operational reports (JSON) indexed by client_id. The payload is sent back to n8n for persistence in the Operational Database (Supabase – RLS by org_id). Pluviometric Agent (Weather Forecast and Risk Analysis) Fetches data from external APIs (INMET, OpenWeather) and produces 3- to 7-day precipitation and wind forecasts. Normalizes data by coordinate, calculates risk indexes, and writes results through n8n to the Operational Database. Executive Reporting Agent (Synthesis and Analytics) This agent does not consume data from other agents directly. Instead, it queries the Supabase databases (Legal and Operational) to retrieve only persisted and audited records. It aggregates, compares, and summarizes these datasets to produce weekly and monthly executive reports, which are then persisted back through n8n. -Workflow Summary Webhook → Gatekeeper → Specialized Agent → n8n → Supabase (Legal/Operational) → Reporting Agent (queries DB) → n8n → Final Storage -Security & Encryption In-Transit Encryption (TLS 1.3) across all services. At-Rest Encryption (AES-256) for databases and object storage. Key Management (Vault or KMS) with rotation policies, RBAC, and per-client segregation. -Hybrid auditing model: DB-side triggers/functions for transactional consistency. App-side logs (Gatekeeper/n8n) for orchestration lifecycle events. All payloads must be JSON-schema-validated (AJV/Zod) and digitally signed (JWS) for data integrity. -Tech Stack n8n self-hosted (Queue Mode, Redis, PgBouncer, JWT auth). Node.js/TypeScript microservices (Gatekeeper + agents as async workers). Supabase PostgreSQL with RLS and schema-per-client isolation. Redis or RabbitMQ as message broker for asynchronous workers. Docker + Kubernetes (autoscaling, blue-green deployments). Prometheus + Loki + Grafana for observability and structured audit logging. Vault or KMS for key management and signing. -Desired Skill Set Experience with n8n (queue mode, webhooks, retries, Wait Nodes). Strong Node.js / TypeScript and PostgreSQL (RLS, JSONB, pgvector) skills. Familiarity with OCR/LLM integration (AWS Textract, OpenAI, Claude, Mistral). Knowledge of Docker/Kubernetes deployments and DevOps pipelines. Strong understanding of data security (TLS, AES, JWT, Vault/KMS, RBAC).