GenAI Engineer | Hungary

Who we are looking for

🎯We’re building agentic systems that do real work in production: tool-using workflows, RAG, structured outputs, guardrails, and measurable quality. If you like shipping and iterating fast - but still care about doing it properly - this is the team.

You’ll work with Solution Architects and LLMOps, but your focus is simple: build the thing, make it reliable, and keep improving it.

Your future role

Agent workflows with LangGraph / LangChain / PydanticAI / CrewAI (routing, retries, fallbacks, timeouts, human-in-the-loop). Tool calling that doesn’t break: APIs, databases, and internal services with clean contracts, predictable behavior, and safe error handling (retries/timeouts, idempotency, rate limiting). Hybrid systems: pre-built agents plus a thin custom orchestration layer (interfaces, policies, guardrails, reuse). MCP integrations: implement MCP servers/clients so models can safely access tools/data (DBs/APIs/files) using least-privilege patterns and audit-friendly logging. RAG + knowledge systems: chunking, embeddings, indexing, retrieval strategies, grounding patterns; vector stacks like Azure AI Search, Pinecone, Redis Vector, Milvus/Chroma; doc ingestion with Azure Document Intelligence. Real use cases: Talk2Data (safe querying + interpretation of enterprise data) Talk2Document (Q&A/summarize/extract/reason over docs with citations/grounding) Evaluation + observability: automated evals (task success, groundedness/relevance, safety, latency, cost), regression suites, and run tracing for debugging using LangSmith / Langfuse / Datadog. Cloud GenAI: Azure OpenAI / Azure AI Foundry + Azure AI Search / Document Intelligence / Content Safety. AWS/GCP equivalents welcome (Bedrock/Vertex AI + search/document pipelines). What we offer

• A global network + a strong regional AI&D community (~160 people).
• A dedicated technical team that’s actively growing and open to experimenting with new tech (when it actually helps).
• Consultancy work, but on real problems with real impact - solutions people use.
• Specialize in specific areas where we are committed to support the learning process and obtaining internationally recognized certificates
• Hybrid way of work
• International environment where personal development and growth are supported and highly encouraged
• Unique opportunity for a career and promotions

🔎What you bring:

• 2+ years as a GenAI Engineer; ideally some NLP background for RAG and Talk2Document work.
• Python-first engineer who can build production APIs/services (FastAPI or similar).
• You’ve built agentic systems with LangGraph / LangChain / PydanticAI / CrewAI (or similar).
• You’re comfortable integrating tools/data safely (timeouts, retries, idempotency, rate limiting).
• You’ve worked with RAG/vector search (Azure AI Search, Pinecone, Redis Vector, Milvus/Chroma).
• You’ve shipped Talk2Data and/or Talk2Document-style solutions (or very similar patterns).
• You care about quality: eval harnesses + regression suites, and you can use LangSmith/Langfuse/Datadog to debug what’s happening.
• Familiar with MCP and secure model-to-tool/data connectivity.
• Bonus: A2A patterns / agent-to-agent coordination.
• Bonus: agent builder frameworks (Azure AI Foundry, AWS Bedrock Agents, Vertex AI Agent Builder, or similar).
• Cloud experience: Azure preferred, AWS/GCP fine.
• EU work permit

About Deloitte

Deloitte is one of the world's leading professional services firms, providing world-class audit and accounting advisory, tax and legal advisory, risk advisory, and consulting services to clients in more than 150 countries, supported by approximately 334,000 professionals globally.