Reliable self-service analytics
I help data teams turn the reporting bottleneck into trustworthy self-service, for people and AI agents alike.
Self-service was supposed to free your analysts from the ad-hoc SQL treadmill. Instead, answers nobody can trust keep the bottleneck firmly in place. The fix is not a smarter model. It is context.
What I work on
Trustworthy answers from your structured data. Governed semantic and context layers, so a business user and an AI agent reach the same number, to the cent.
Turn unstructured documents into structured data you can actually trust.
The semantic backbone that connects your data, your documents, and your agents.
Proof of work
End-to-end analytics platform
dlt to dbt (Kimball, enforced contracts) to a Cube semantic layer to self-service BI, with governed metrics exposed to LLM agents via MCP. The same number, to the cent, across four surfaces: Cube, Metabase, raw SQL, and an AI agent. Terraform, CI/CD, 13 ADRs.
Grounded agent on a knowledge graph
An LLM agent that answers only from a Neo4j knowledge graph, with a mandatory source on every node, explicit refusals out of scope, and a two-level evaluation suite over 28 bilingual cases.
Featured writing
Reliable self-service is a context problem, not a model problem. The anatomy, layer by layer, built on a zero-cost stack, gaps included.
Read the anatomy →More deep-dives coming: procedural skills and offline evals.
I'm a Data & AI Engineer working at the intersection of Analytics Engineering and Agentic AI.
Before that, I spent four years on the Finance & Accounting modules of a multi-currency ERP running in production at Médecins Sans Frontières: several thousand users, 70+ countries, monthly and annual closings, reconciliations, exchange rate handling. An integrity issue could block a closing or delay payments to field staff and suppliers. That is where my focus on reliability comes from.
I build systems where business users, analysts, and AI agents rely on the same trusted metrics and reach the same answers.
My work focuses on making analytics systems robust by making context explicit through semantic layers and metrics governance.
On top of that, I explore how document AI and knowledge graphs can extend structured analytics systems with reliable context for retrieval and reasoning.
Currently open to permanent roles in France as a Data & AI Engineer, Analytics Engineer, or AI / Applied AI Engineer.
Gaël Mukunde