GenAI Engineer; LLM Azure

Project description

Expert-level engineer designing and building LLM applications and agentic workflows on Azure: RAG pipelines, multi-agent orchestration, MCP tool integration, evaluation, guardrails and cost control. Sets the GenAI engineering standards for the account and is accountable for solution quality against agreed evaluation thresholds.

• Build LLM applications and agent workflows with Lang Graph / Lang Chain / Semantic Kernel on Azure OpenAI (AWS Bedrock / GCP Vertex a plus).
• Design RAG pipelines: document ingestion, chunking strategies, embeddings, vector search (Azure AI Search, FAISS, Pinecone), hybrid retrieval and re-ranking.
• Integrate enterprise tools and context via Model Context Protocol (MCP) servers and function / tool calling; design multi-agent systems.
• Engineer, version and test prompts; run offline and online evaluations (accuracy, grounding, hallucination rate) and fine-tuning where justified.
• Implement responsible-AI guardrails (content filters, PII redaction, policy checks) and LLM observability (tracing, token accounting, latency).
• Optimise token and inference cost (model routing, caching, batching); report against cost budgets.
• Document prompts, evaluation results and model choices; mentor engineers adopting GenAI patterns.

• Bachelor's degree in Computer Science, Engineering, Information Systems or a related field, or equivalent practical experience.
• 10+ years in software or data engineering, of which 3+ in machine learning / NLP and 2+ delivering production LLM applications (RAG, agents) on Azure OpenAI or equivalent.
• Expert Python;
  Lang Graph, Lang Chain and/or Semantic Kernel; asynchronous, event-driven application design.
• Azure OpenAI, Azure AI Search, Azure AI Foundry / ML, Functions, Container Apps / AKS, Key Vault, Application Insights.
• RAG design: chunking, embeddings, vector databases (Azure AI Search, FAISS, Pinecone), hybrid search, re-ranking.
• MCP server / client development, function calling, multi-agent orchestration and state management.
• LLM evaluation (RAGAS, promptfoo, Azure AI evaluation or equivalent), fine-tuning (LoRA / PEFT) and model selection.
• Guardrails (Azure AI Content Safety, NeMo Guardrails or similar), LLM observability (Lang Smith, Open Telemetry), Fin Ops for inference.

• AWS Bedrock or GCP Vertex AI.
• Semantic caching, model distillation, small-language-model deployment.
• Azure AI Engineer Associate (AI-102) or equivalent.
• Experience in financial services, sovereign wealth / investment holding or other regulated enterprise environments.
• Experience working with distributed teams (onsite UAE with nearshore India / offshore Poland squads).