Job Description:

Role Overview:

The Platform & Orchestration Engineer is a critical new role responsible for designing, building, and operating the durable execution layer that underpins all AI agent workflows in the Agentic AI ERP Platform. This role owns the Restate runtime deployment, Virtual Object patterns, awakeable lifecycle management, audit integration, and the workflow-as-data engine that enables runtime-configurable business processes.

This is a foundational hire. The role bridges distributed systems engineering with AI agent architecture, ensuring that multi-step ERP workflows (spanning hours to weeks) survive crashes, guarantee exactly-once execution, and maintain SOX-grade audit trails.

Key Responsibilities:

Orchestration Platform (Primary):

- Deploy, configure, and operate Restate clusters in production Kubernetes environments (multi-node, cross-AZ replication, S3 snapshots).

- Design and implement Virtual Object patterns for business entities (invoices, purchase orders, vendors) - each a 'living' durable object with its own state and lifecycle.

- Build awakeable patterns for human-in-the-loop approval workflows that can suspend for days and resume on external signals.

- Implement durable timer patterns for non-event detection (overdue approvals, missing payments, SLA breaches).

- Design idempotency strategies per ERP system to guarantee exactly-once writes for financial transactions.

- Monitor and optimize Restate performance using the embedded SQL engine, UI, and CLI tooling.

- Manage service versioning and deployment draining to ensure zero-downtime upgrades of workflow logic.

Workflow Engine & Configurability:

- Design and build the workflow-as-data engine that enables runtime-configurable business processes stored as JSON/YAML definitions in PostgreSQL.

- Create parameterized workflow templates for each Rimini Solution (Finance, Procurement, Supplier, Expense, Support) with customer-adjustable decision points.

- Implement the generic Workflow Engine Virtual Object that interprets workflow definitions at runtime - approval thresholds, routing rules, escalation timing, autonomy tier per step.

- Build the foundation for a future visual workflow designer by ensuring all workflow logic is externalized from compiled code.

Governance & Audit Integration:

- Own the OPA (Open Policy Agent) policy-as-code layer - write, test, and deploy Rego policies for risk-tiered autonomy (LOW/MEDIUM/HIGH).

- Build the dual-stream audit bridge: operational logs (OpenTelemetry - Jaeger/Prometheus) and compliance logs (structured records - PostgreSQL with 7-year retention).

- Ensure every ctx.run() step in Restate handlers produces a corresponding structured audit record capturing who, what, when, authorization decision, and before/after state.

- Implement the circuit breaker hierarchy (rate limiter - cost governor - error rate breaker - objective breaker - kill switch).

Proactive Agent Infrastructure:

- Implement the four-layer proactive architecture: Sensing (events/timers/MCP Sampling) - Governance (OPA) - Execution (Restate + reactive agents) - Audit (dual-stream logging).

- Build the 'Self-Waking Virtual Object' pattern enabling millions of monitoring agents at near-zero cost through durable sleep.

- Integrate ERP event sources (SAP Event Mesh, Oracle EBS Business Event System) with deterministic filter pipelines.

- Implement trigger recursion detection and prevention ('Did I cause this?' deduplication checks).

Required Experience:

- 6+ years of software engineering experience with strong Java or Kotlin proficiency.

- 3+ years working with distributed systems - event sourcing, state machines, message queues, or workflow engines.

- Production experience with at least one durable execution or workflow orchestration platform (Temporal, Restate, AWS Step Functions, Cadence, Conductor, Camunda, or similar).

- Hands-on Kubernetes experience - deploying StatefulSets, managing persistent storage, configuring operators.

- Experience with event-driven architectures (Kafka, RabbitMQ, or cloud-native event services).

- Understanding of exactly-once semantics, idempotency patterns, and distributed transaction strategies (Saga, compensating transactions).

- Experience implementing audit logging and compliance controls in regulated environments (SOX, SOC 2, GDPR) is strongly preferred.

Required Technical Skills:

- Java 21 / Quarkus: Core platform language and framework - must be proficient.

- Restate or Temporal: Durable execution platforms - production experience with either; willingness to become Restate expert.

- Kubernetes: Cluster operations, Helm charts, operators, StatefulSets, persistent volumes, pod affinity/anti-affinity.

- PostgreSQL: Audit tables, structured compliance logging, query optimization.

- OPA / Rego: Policy-as-code authoring and testing (or strong willingness to learn).

- OpenTelemetry: Distributed tracing instrumentation, span creation, context propagation.

- Event Streaming: Kafka or equivalent - producing/consuming events, partitioning, consumer groups.

- RocksDB: Understanding embedded storage (Restate's internal store) - helpful but not required.

Preferred Skills:

- Experience with ERP systems (SAP, Oracle EBS, JD Edwards) - even basic familiarity is valuable.

- DBOS Transact experience (PostgreSQL-based durable execution - our fallback option).

- Workflow engine design - building configurable workflow definitions, visual builders, or BPM-style tools.

- AI/LLM integration - understanding agent loops, tool calling patterns, non-deterministic execution flows.

- Redis - caching patterns for permission manifests and OPA policy evaluation results.

- Grafana / Prometheus - building operational dashboards for workflow health monitoring.

- Air-gap deployment experience - self-hostable architectures without cloud dependencies.