09/23/2026 updated
100 % available
AI Engineer | Production LLM & Agentic Systems | GraphRAG · Knowledge Graphs · Neo4j | A2A · MCP
Timișoara, Timiș, Romania
Only remote
BSc Computer Science, MSc Advanced Math Algorithms, AWS AI Practitioner, PCEP/PCAP/PCPP1 PythonAbout me
AI Engineer combining 11 years of enterprise software experience with hands-on Agentic AI, RAG/GraphRAG and Knowledge Graph engineering. I build reliable AI systems with Python, LangGraph, MCP, Neo4j, evaluation and observability—bridging AI experimentation and production engineering.
LLaMAGCPLlamaIndexLangChainPython (Programming Language)PostgreSQLNeo4jNode.jsRabbitMQRedisPrometheusTwilioTypeScriptWebSocketWebRTC
AI Engineer building production-oriented Generative AI, Agentic AI, Knowledge Graph, Context Graph, GraphRAG, RAG, observability, voice AI, and multi-agent systems.
I design reliable LLM applications with deterministic orchestration, persistent state and memory, MCP/A2A tools, HITL approvals, evaluation, tracing, tenant isolation, and fault-tolerant workflows.
My flagship Neo4j/RDF Knowledge Graph and Sales Context Graph platform combines hybrid and bitemporal retrieval, multi-hop traversal, BM25/RRF, cross-encoder reranking, GAT/GCN re-scoring, entity resolution, OWL/OWL-RL reasoning, SPARQL, SHACL validation, semantic caching, adaptive routing, provenance, contradiction detection, and governed decision traces.
It exposes versioned capabilities through authenticated Streamable HTTP MCP and supports enterprise-safe writes with entitlements, idempotency, optimistic concurrency, dry-run previews, human approval, immutable receipts, and compensation. The Context Graph records cases, agent runs, evidence, policies, alternatives, observations, decisions, structured rationale, and integrity-verified inference context without storing hidden chain-of-thought.
Results: 0.940 RAGAS faithfulness on a 39-question archived aerospace evaluation; 30/30 golden graph-fact evaluations; 26× faster warm embedding cache, from 2,859 ms to 109 ms. Validation includes LLM-as-a-Judge, golden datasets, load/failure testing, OpenTelemetry, security exercises, Docker, Minikube/Kubernetes, Terraform tests, and Checkov.
I bring 10+ years of enterprise software engineering experience, including high-availability telecom platforms at Alcatel-Lucent, distributed systems, APIs, event-driven architecture, security, observability, CI/CD, and live Cloud Run deployments.
Tech stack: Python, TypeScript, FastAPI, React, Node.js, Next.js, LangGraph, LangChain, LlamaIndex, Haystack, AutoGen, CrewAI, Google ADK, Mastra, PydanticAI, DSPy, OpenAI Agents SDK, Gemini, GPT-4o, Claude, Llama, Neo4j, Cypher, RDF, OWL, OWL-RL, SPARQL, SHACL/pySHACL, Datalog, PageRank, Leiden, TransE, GNN, PyTorch, sentence-transformers, BM25, RRF, PostgreSQL, pgvector, OpenSearch, Elasticsearch, Qdrant, Pinecone, Chroma, Milvus, Weaviate, FAISS, Redis, Kafka, RabbitMQ, gRPC, Envoy, WebSocket, WebRTC, Twilio, Deepgram, Docker, Kubernetes, Terraform, GCP, Cloud Run, Vertex AI, AWS SageMaker, GitHub Actions, OpenTelemetry, Prometheus, Grafana, VictoriaMetrics, Jaeger, Langfuse, LangSmith, RAGAS, MCP, FastMCP, A2A, HITL, JWT, OIDC/PKCE, ABAC, Semgrep, Bandit, pip-audit, ServiceNow.
Experience
Professional experience10+ years
Freelancer since2025
Languages
EnglishNative speakerFrenchBasic knowledge
Project history
Designed and built production-oriented AI systems across Agentic AI, RAG/GraphRAG, Knowledge Graphs and LLM evaluation.
• Developed Python/FastAPI backends and multi-agent workflows using LangGraph, AutoGen, CrewAI and Google ADK, with persistent state, tool calling, HITL and structured outputs.
• Built RAG/GraphRAG pipelines combining vector search, BM25, RRF, cross-encoder reranking, multi-hop graph retrieval and grounded generation.
• Designed Neo4j Knowledge Graph and semantic architectures using RDF/OWL, SHACL, SPARQL, entity resolution, provenance and temporal data.
• Implemented MCP and A2A agent communication, governed tool access, authentication and API integrations.
• Built LLM/RAG evaluation pipelines using RAGAS, LLM-as-a-Judge, golden datasets and retrieval metrics including Recall@K, MRR and nDCG.
• Implemented AI observability with OpenTelemetry, Langfuse, Prometheus and Grafana.
• Built real-time Voice AI workflows integrating Twilio, WebRTC, Deepgram STT and TTS.
• Containerized and deployed AI services using Docker, Kubernetes, CI/CD and cloud platforms, primarily GCP.
• Implemented guardrails, PII controls, tenant isolation, testing and reliability mechanisms for production-oriented AI systems.