We are seeking an AI Engineer to lead the development of our Pro-Code autonomous agent ecosystem. You will be architecting the foundational orchestration, safety, and serving layers of our AI platform.
The ideal candidate is an expert in navigating the Open Source AI landscape, capable of replacing proprietary black-box services with high-performance, self-hosted alternatives. You will be responsible for building complex multi-agent workflows, implementing rigorous governance, and delivering high-fidelity user interfaces.
This role will be fully on site in Abu Dhabi and will be offered as 12 month extendable contracts.
Key Responsibilities
1. Advanced Agentic Orchestration
- Design and implement complex, stateful multi-agent systems using LangGraph or CrewAI.
- Build Agentic RAG pipelines that utilize LlamaIndex or Haystack or similar for advanced reasoning over structured and unstructured datasets.
- Implement custom tool-calling logic and Model Context Protocol (MCP) integrations to connect agents with internal databases and APIs.
2. AI Safety & Governance (The Guardrails Layer)
- Architect a unified Safety Perimeter using Guardrails AI or NVIDIA NeMo Guardrails or similar.
- Develop real-time monitoring and observability for LLM inputs/outputs using Arize Phoenix, LangSmith (OSS version), or MLflow or similar.
3. High-Performance Inference & LLMOps
- Optimize and deploy open-source models (Llama 3, Mistral, Qwen or similar) using high-throughput serving engines like vLLM, TGI (Text Generation Inference), or Ollama.
- Manage model lifecycles and experimental tracking using MLflow or BentoML or similar.
- Build and maintain scalable AI infrastructure on Kubernetes (using KubeRay or SkyPilot) for distributed inference and fine-tuning.
4. Full-Stack UI & Custom Interfaces
- Build low-latency, streaming-first user interfaces using React and the Vercel AI SDK (Open Source components) or similar.
- Develop internal human-in-the-loop (HITL) dashboards that allow subject matter experts to intervene, approve, or correct agent actions in real-time.
