Lokesh Sharma

Blog

Notes on knowledge graphs and applied AI.

⌘K

Semantic & Standards

Ontology engineering, semantic web modeling, SHACL, and RDF validation.

Series · 3 parts

Agile Ontology Engineering

Ontology EngineeringData ModelingKnowledge Graphs
1
From Requirements to OWL20 min
→2
SHACL Quality Gates in CI/CD12 min
→3
Persistent IRIs, Semantic Releases, and the Maintenance Loop11 min
→

Choosing Your Upper Ontology

A pragmatic guide to selecting, aligning, and adopting top-level ontologies: BFO, DOLCE, UFO, ISO 15926, gist, CCO, and IOF Core across Energy, Supply Chain, Automotives, and Finance without incurring semantic complexity bloat.

14 min read
Upper OntologiesKnowledge GraphsSemantic Architecture

Why I Start with the Ontology

A framework for trust in the era of AI. Why schema-on-read failed us, why LLMs demand deterministic grounding, and a roadmap to the engineering systems documented in this portfolio.

2 min read
Knowledge GraphsSemantic WebApplied AI

Applied AI in Production

MLOps, monitoring and drift, inference optimization, and the economics of serving models.

Series · 4 parts

End-to-End MLOps

MLOpsProduction MLApplied AI in Production
1
Why Models Fail to Ship13 min
→2
Building the Model18 min
→3
Shipping and Running It20 min
→4
Monitoring and Surviving Silent Failure16 min
→

Data Engineering and Architecture

Pipelines, cloud architectures, systems engineering, and semantic databases.

Draft

Ingestion is not import

"We'll just load the documents" is where document-AI projects go to die. Loading is a one-shot verb; real ingestion is a replayable, idempotent, provenance-stamped pipeline, because you will process the same corpus a dozen times before you're done.

6 min read
Data EngineeringPipelinesDocument AI

Synergy of AI and Knowledge Graphs

MCP, governed agents, GraphRAG, and neuro-symbolic data products.

Series · 2 parts

Governed Agent Writes

Synergy of AI and Knowledge GraphsAI AgentsMCP
1
When an AI Agent Writes to Your System of Record18 min
→2
Building the Governed Write Path13 min
→
Draft

Giving RAG a sense of time

Vector search finds similar text; it cannot tell you how things relate or when they were true. Pairing an ANN index with a temporal knowledge graph gives document AI both fuzzy entry and precise, time-aware structure.

6 min read
Synergy of AI and Knowledge GraphsGraphRAGRAG
Draft

The entity linking flywheel

From raw documents to governed graph facts: the NER ladder, two-stage disambiguation, NIL detection, and why the knowledge graph itself is the best feature your linker will ever have.

5 min read
Knowledge Graph ConstructionEntity LinkingNER