The Umanitek Guardian Agent protects people from deepfakes, identity abuse, and coordinated harmful narratives by combining Decentralized Knowledge Graphs (DKG) with graph-native risk intelligence. If you’re excited by building resilient data infrastructure and operational systems for agents that reason over verifiable evidence, not just predict from black-box signals, you’ll feel right at home.
As a AI & Knowledge Graph Engineer, you’ll work at the frontier of neuro-symbolic AI, connecting LLM-assisted analysis with structured graph reasoning and provenance. You’ll be responsible for building and evolving our core product – the Umanitek Guardian pipeline and agent network – an AI system that protects humans from deepfakes and impersonation by grounding intelligence in RDF knowledge graphs. Your focus will be to design and apply hybrid reasoning capabilities, combining SPARQL-driven graph logic, risk scoring, entity resolution, embedding similarity, and media analysis into one coherent threat-detection system.
Responsibilities:
* Develop and optimize hybrid reasoning systems combining Knowledge Graphs, LLMs, embedding search, and symbolic reasoning
* Integrate media-analysis capabilities (image/video processing, OCR, similarity signals) for harmful content discovery and tracking
* Integrate perceptual hashing and similarity search algorithms for image/video discovery
* Build and evolve the Guardian stack (RDF, SPARQL, provenance modeling, schema/ontology design) for neuro-symbolic reasoning
* Build and test MCP tools enabling Guardian’s AI to safely query other Guardian Agents on the Guardian Network
* Measure and improve risk scoring, graph reasoning quality, and end-to-end detection performance
* Ensure scalability, data quality, and operational reliability of the Guardian knowledge base and orchestration pipeline
Technical experience that is most relevant to us:
* Strong background in data engineering and production ETL/pipeline systems
* Experience with knowledge graphs, particularly RDF, SPARQL, Blazegraph, and reasoning approaches
* Experience with LLMs, embeddings, or reasoning systems (LangChain, Semantic Kernel, etc.)
* Understanding of perceptual hashing (pHash, aHash, dHash) and multimedia similarity search
* Understanding of multimedia threat analysis, including similarity methods and practical approaches to image/video signal matching
Tech Stack and Tools: Python, Rust, Nodejs, PyTorch, HuggingFace, LangChain, Qdrant/Weaviate/Milvus, OpenAI/Gemini API, RDFLib, GraphDB, SPARQL, FAISS, TensorFlow, FastAPI, Docker, GitHub Actions, MCP
We offer a market‑aligned salary, participation in employee stock option pool and of course growth into companies’ leadership positions.
#J-18808-Ljbffr