Senior specialist · AI / data / analytics

Building useful AI systems with the governance visible.

Hands-on technical depth with delivery leadership across specialist agents, secure tool interfaces, enterprise analytics and human-accountable delivery.

Explore the work See the operating model

Human-accountable delivery loop

AIorchestrator

01 Evidence

02 Specialists

03 Review

04 Approval

Reason → verify → approve → publish

01Agent orchestration

02Secure MCP

03Data & analytics

04Human governance

Selected work

Cases built around the hard parts.

Architecture, constraints, ambiguity and verification—not just polished outcomes.

01 · Agent orchestrationOperational

SAP HANA regression readiness

Automated six-phase regression checks for SAP HANA analytics — specialist-gated findings, controlled reporting.

Read the case study

02 · Secure MCPOperational

Secure MCP operations

A governed MCP boundary that turned live network telemetry into bounded, verified fixes.

Read the case study

03 · Deterministic automationOperational

Evidence-driven home automation

Deterministic climate control with AI-built evidence — and an honest verdict on whether it helped.

Read the case study

04 · Agent orchestrationOperational

AI delivery operating model

The pipeline behind this site: bounded agent roles, deterministic gates, human consent.

Read the case study

05 · Agent orchestrationOperational

Agentic wealth operations

Cost-tiered agents keeping self-hosted ledgers current under a written operations contract.

Read the case study

06 · Data governanceOperational

Governed personal health data

A health data platform with no database port — provenance-first ingestion, pinned metrics, gated writes.

Read the case study

07 · Agent orchestrationOperational

Agentic coaching with human gates

Agents that plan and review my training — judgment separated from execution, nothing applied unverified.

Read the case study

View all work

Operating model

AI delivery is a system, not a prompt.

Work moves through bounded specialists, explicit evidence, independent review and human approval. Model choice follows the task—not habit.

Explore the model

01
Frame
Define the decision, evidence and stop conditions.
02
Route
Choose tools, specialists and model capacity deliberately.
03
Verify
Test outputs, read changes back and preserve uncertainty.
04
Approve
Keep acceptance and external action human-accountable.

The working principle

“Make the reasoning useful. Make the boundary obvious.”

Connect on LinkedIn