A small applied-AI lab publishing reference implementations of the patterns I run at exec engineering altitude.
- AI governance and Responsible AI — ISO/IEC 42001 Annex A controls, EU AI Act Articles 12–15, NIST AI RMF, UK AI Cyber Code, turned into runtime-enforced policies and evidence-generating logs.
- Agentic AI harnesses — autonomy budgets, blast-radius controls, regulator-legible audit trails. The shape of the patent I filed at A.P. Moller–Maersk, generalised to non-customs domains.
- LLM control plane — DLP, PII redaction, prompt-injection defence, ephemeral pods, policy-as-code, observability, evals.
- Graph RAG and ontology-based retrieval — three-layer semantic ontologies over public regulatory corpora (EUR-Lex AI Act, GDPR, DORA, NIS2).
- Every repo is a public reference implementation. No employer IP, no client data, no proprietary architecture. Synthetic and public datasets only.
- Each repo is paired with a long-form essay on themindfulcto.com explaining the design choices and the trade-offs that produced them.
- We delete more than we add. The portfolio shrinks before it grows.
- Plain English. No badge soup. No "revolutionary AI". The code is the claim.
@venkat-uk — Engineering Director leading AI Supported Development, AI Governance, and Responsible AI at a UK retail-data platform. Previously Head of Technology for a logistics platform across 130 countries. Patent application pending on agentic AI for regulated workflows.
Unless noted otherwise inside individual repos, code in this organisation is Apache 2.0 and documentation is CC BY-SA 4.0.