remoteonsite
Z2 Full Stack AI Engineer - NTT DATA
AI Engineer
We're looking for an AI Engineer focused on designing and building scalable technical solutions. This mid level role requires 4+ years of relevant experience.
About the role
- Design and implement multi-agent orchestration pipelines using LangGraph state machines with conditional routing, self-healing loops, and sub-agent hierarchies
- Build and maintain a hybrid retrieval system combining Neo4j knowledge graphs (Cypher) with Qdrant vector search and sentence-transformer embeddings
- Develop FastAPI backend services with async I/O (asyncpg, Motor), Pydantic v2 models, and JWT/RBAC security
- Build React 18 / TypeScript frontend features including visual graph editors (ReactFlow), real-time SSE streaming, and observability dashboards (Recharts)
- Integrate enterprise protocols: A2A (JSON-RPC), MCP (Model Context Protocol), AG-UI (Server-Sent Events)
- Connect to ALM/PLM tools via REST and GraphQL APIs with OAuth2/SSO authentication (SAML, OIDC)
- Implement dual-tier agent memory (Redis L1 + mem0 L2) with semantic recall and context consolidation
- Build DAG-based workflow compilation with topological sorting, cycle detection, and Ray-based parallel execution
- Deploy and operate containerized services using Docker, Azure Bicep IaC, and CI/CD pipelines (Azure Pipelines / Jenkins)
- Serve and manage on-prem LLM models via vLLM alongside cloud-hosted models (OpenAI-compatible APIs)
- Python 3.10+ — production-grade FastAPI, Pydantic v2, async/await patterns
- LangGraph / LangChain — state machines, tool calling, agent orchestration
- LLM engineering — prompt design, token budgeting, multi-model routing, self-healing retry loops
- PostgreSQL (JSONB, asyncpg) or MongoDB (Motor) — async data persistence
- Neo4j — Cypher queries, entity-relationship graph modeling, traversal algorithms
- Vector databases (Qdrant, Pinecone, or Weaviate) — embedding pipelines, semantic search, hybrid retrieval
- RAG architecture — chunking strategies, re-ranking, token budget management - Sentence-transformers or equivalent embedding models
- React 18 + TypeScript + Vite — component architecture, hooks, lazy loading
- Server-Sent Events (SSE) or WebSocket — real-time streaming UIs
- ReactFlow / XYFlow — interactive node-based graph editors
- State management (Zustand or Redux) + server-state (TanStack Query)
- vLLM or TGI — on-prem LLM model serving and optimization
- mem0 or similar — persistent agent memory systems
- A2A / MCP / AG-UI — emerging agentic interoperability protocols
- GraphQL — schema design, client integration
- OAuth2 / SAML / OIDC — enterprise SSO integration
- Ontology modeling — entity taxonomies, relationship schemas for engineering domains
- Pytest — async test patterns, fixture-based testing
- 4+ years professional software engineering experience
- 2+ years working with LLMs, agent frameworks, or ML systems in production
- Strong async Python and modern React/TypeScript skills