Build Production Multi-Agent AI System

We are looking for a senior (Principal / Staff+ level) engineer to architect, build, secure and operate production-grade autonomous multi-agent AI systems. This is a long-term engagement on real production systems with real uptime requirements, not a chatbot, demo or no-code automation.

WHAT YOU WILL BUILD
• Multi-agent systems with planner, executor, supervisor, worker and critic agents (LangGraph or custom)
• Event-driven, durable workflows that run for hours or days and recover from any failure (Temporal, Celery, Kafka, NATS)
• Inference and model-routing infrastructure optimized for cost and latency (vLLM, Ray, OpenAI and Anthropic APIs)
• Persistent memory and advanced retrieval: Postgres, Redis, pgvector, Neo4j, Graph RAG, hybrid search
• Tool and enterprise integrations via MCP and APIs (Salesforce, SAP, ServiceNow, Oracle, Workday)
• Human-in-the-loop controls: escalation, approvals, confidence thresholds and override
• Evaluation, monitoring and audit logging for every agent decision (LangSmith, OpenTelemetry, Prometheus, Grafana)
• Deployment on Docker, Kubernetes and Terraform across AWS, Azure or GCP

REQUIREMENTS
• Proven experience building and operating autonomous or multi-agent systems in real production
• Ability to own the full system: architecture, infrastructure, deployment, security, evaluation and reliability
• Strong distributed systems skills: idempotency, retries, replay, backpressure, fault tolerance
• Security-first approach: prompt injection, tool abuse, memory poisoning and access control
• Deep async Python (asyncio, concurrency, queues)
• Experience setting and meeting reliability targets (SLOs) for latency, uptime and cost
• Production integration experience with at least two of Salesforce, SAP, ServiceNow, Oracle or Workday
• Able to work independently without a technical lead

PLEASE DO NOT BID IF
Your main experience is chatbot wrappers, simple RAG apps, Streamlit demos, no-code workflows or prompt engineering only.

PROJECT STAGES
1. Paid 90-minute live architecture review
2. Paid 3–4 hour live build of a working multi-agent system (it must run)
3. Review call on a production system you have built
4. Live security and reliability review of your design
5. Ongoing long-term work

HOW TO BID
Bids without these will not be considered:
1. Your GitHub and LinkedIn links
2. The most impressive autonomous system you have shipped, your role, and what runs in production today
3. The hardest production failure you fixed and how you prevented it from happening again
4. Your hourly rate and weekly availability

We are looking for a senior (Principal / Staff+ level) engineer to architect, build, secure and operate production-grade autonomous multi-agent AI systems. This is a long-term engagement on real production systems with real uptime requirements, not a chatbot, demo or no-code automation.

WHAT YOU WILL BUILD
• Multi-agent systems with planner, executor, supervisor, worker and critic agents (LangGraph or custom)
• Event-driven, durable workflows that run for hours or days and recover from any failure (Temporal, Celery, Kafka, NATS)
• Inference and model-routing infrastructure optimized for cost and latency (vLLM, Ray, OpenAI and Anthropic APIs)
• Persistent memory and advanced retrieval: Postgres, Redis, pgvector, Neo4j, Graph RAG, hybrid search
• Tool and enterprise integrations via MCP and APIs (Salesforce, SAP, ServiceNow, Oracle, Workday)
• Human-in-the-loop controls: escalation, approvals, confidence thresholds and override
• Evaluation, monitoring and audit logging for every agent decision (LangSmith, OpenTelemetry, Prometheus, Grafana)
• Deployment on Docker, Kubernetes and Terraform across AWS, Azure or GCP

REQUIREMENTS
• Proven experience building and operating autonomous or multi-agent systems in real production
• Ability to own the full system: architecture, infrastructure, deployment, security, evaluation and reliability
• Strong distributed systems skills: idempotency, retries, replay, backpressure, fault tolerance
• Security-first approach: prompt injection, tool abuse, memory poisoning and access control
• Deep async Python (asyncio, concurrency, queues)
• Experience setting and meeting reliability targets (SLOs) for latency, uptime and cost
• Production integration experience with at least two of Salesforce, SAP, ServiceNow, Oracle or Workday
• Able to work independently without a technical lead

PLEASE DO NOT BID IF
Your main experience is chatbot wrappers, simple RAG apps, Streamlit demos, no-code workflows or prompt engineering only.

PROJECT STAGES
1. Paid 90-minute live architecture review
2. Paid 3–4 hour live build of a working multi-agent system (it must run)
3. Review call on a production system you have built
4. Live security and reliability review of your design
5. Ongoing long-term work

HOW TO BID
Bids without these will not be considered:
1. Your GitHub and LinkedIn links
2. The most impressive autonomous system you have shipped, your role, and what runs in production today
3. The hardest production failure you fixed and how you prevented it from happening again
4. Your hourly rate and weekly availability

Start your bid with the word "DETERMINISTIC" so we know you read the full description.

Start your bid with the word "DETERMINISTIC" so we know you read the full description.