# Andrew Stevens - Long-Form Profile

> Advisor on AI governance, cloud architecture, and security. Open to technology leadership and transformation roles. Author of GATE, an open control framework for AI agents, and of Tollgate, Arkhe, and sql-guard. Co-Founder, Sakura Sky. Board Member, Distinct Star.

## Identity

- Name: Andrew Stevens
- Locations: Amsterdam, Netherlands (EU office); New York, USA. Sakura Sky's US head office is in San Francisco.
- Website: https://www.andrewjstevens.com
- Email: andrew@andrewjstevens.com
- LinkedIn: https://www.linkedin.com/in/andrewjstevens
- Mastodon: https://mastodon.online/@andrewjstevens
- GitHub: https://github.com/andrewjstevens
- GitHub (GATE): https://github.com/deterministic-agents
- PGP Fingerprint: F9B3 CECE F5D4 9ABD 682B 388B 17F8 4E48 609C 3BBB

## Summary

Andrew Stevens works with leadership and engineering teams on AI governance, cloud architecture, and security, drawing on two decades as a founder, CTO, and CISO building cloud platforms and security programmes in regulated and high-growth businesses. In AI, he helps organisations design systems they can govern: agent architectures where it is clear what each agent is allowed to do, under whose authority, what it touched, and how to stop it mid-task.

Most teams he speaks with can get an agent working quickly. Far fewer can answer those questions with evidence. In his experience that gap holds enterprise AI back more often than model capability does, and since 2025 it has been the centre of his work.

Behind that sits two decades in founder, CTO, CIO, and CISO roles across four continents, in regulated wealthtech in Singapore, blockchain asset succession, industrial systems on mine sites, and global media and travel platforms. The sectors differed; the recurring work was deciding what a system may do and being able to show afterwards that it did only that.

Across that time he has led engineering, product, and security teams through three market shifts: the move to cloud, the move to mobile, and now AI, building teams that could adopt each new platform quickly while keeping control of what they already ran.

## Services

### AI governance and risk
Control frameworks, operating models, and audit evidence aligned to NIST AI RMF, ISO/IEC 42001, and US and EU regulatory expectations.

### Agent and AI platform architecture
Agent identity and authority, policy-as-code, runtime guardrails, observability, and spend control.

### Cloud architecture and platform strategy
Cloud and data platform architecture across AWS, Google Cloud, and Azure, including multi-cloud and sovereign cloud decisions, platform engineering and SRE maturity, and cost.

### Security leadership
Security strategy and programmes for cloud-native and regulated businesses: zero-trust architecture, cloud security posture, threat modelling, AI security, and compliance readiness across GDPR, ISO 27001, and sector regulation.

### Reviews and technology due diligence
Independent reviews of AI systems moving from pilot to production, and of cloud platforms and security programmes ahead of scale or audit. For investors and acquirers: pre-deal technology diligence across cloud cost and architecture, security posture, platform scalability, and AI readiness, with a findings report and a 100-day plan.

### Board and executive advisory
AI strategy, cloud and cyber risk, risk appetite, and investment.

## Engagement formats

- Fractional or interim: a set monthly commitment in a technology leadership role, leading AI, cloud, or security transformation alongside existing teams, or building the function where none exists yet.
- Contract: a defined piece of work, such as an architecture review, a security assessment, or a control framework, with a written deliverable.
- Advisory and board roles.
- Speaking and executive briefings.

Also open to speaking, investment conversations, and pro-bono work in human rights, privacy, and social action.

Engagements are independent of Sakura Sky and Distinct Star. Any relationship is disclosed before work starts, and no follow-on work is referred without the client's consent.

Working style: inside existing teams, in their design documents and repositories, with written findings, decision records the engineers own, and a handover that leaves the team running the work without him.

## Selected Work

Published case studies from engagements Andrew Stevens led or oversaw at Sakura Sky. References and further detail are available on request, under NDA where needed.

- [Xapo Bank: GenAI adoption with governance built in](https://www.sakurasky.com/case-studies/innovation-at-speed-how-xapo-bank-achieved-genai-adoption-in-just-8-weeks/) (banking, AI and governance)
- [API governance framework for SkillsFuture Singapore](https://www.sakurasky.com/case-studies/ssg-api-governance/) (public sector, governance and security)
- [Orbx: immersive simulation on Google Cloud](https://www.sakurasky.com/case-studies/orbx-gke-bigtable/) (simulation, cloud platform)
- [Migrating a live real-time communications platform from AWS to Google Cloud](https://www.sakurasky.com/case-studies/11sight-gcp-migration/) (communications, cloud migration)
- [Starting the San Francisco Giants' move to Google Cloud](https://www.sakurasky.com/case-studies/sf-giants-sports-data-platform/) (sport, data platform)
- [Underground camera feeds in the browser for APC Global Systems](https://www.sakurasky.com/case-studies/apc-global-ip-cameras/) (mining, industrial)
- [All case studies](https://www.sakurasky.com/case-studies/)

## Open Source

Specifications published together with the code that implements them, under open licences, with versioned releases, conformance checks, and documented scope limits. Where a control is only partly automatable, the documentation says so. These are early projects with a single maintainer, published as reference designs rather than adopted standards; Tollgate is not yet production-ready.

### GATE - Governed Agent Trust Environment (v1.4, MIT / CC BY 4.0)
Twenty controls across four layers for governing agent runtimes: identity and integrity, runtime enforcement, observability and forensics, and orchestration. Agent runtimes never call tools or memory directly; side effects pass through enforcement points that emit verifiable evidence. Ships JSON Schema contracts, OPA/Rego policy and invariant bundles, Python and Rust reference libraries, a cross-language differential property suite, a conformance runner, and operational runbooks. Informative mappings to NIST AI RMF, ISO/IEC 42001, OWASP AISVS, MITRE ATLAS, and NIST SSDF. Ten repositories. Explicitly not a product and not a hosted service.
URL: https://github.com/deterministic-agents / https://deterministicagents.ai

### Tollgate (Rust, MIT)
Open-source AI gateway and spend-control proxy for LLM providers. Prices every request by token, reserves budget before the call, and returns a hard 402 before the request reaches the provider. Per-key and global limits, tokens-to-cost accounting, an append-only Postgres ledger, hot counters in Valkey, a read-only console, and Terraform for Cloud Run. Anthropic and Vertex adapters. Pre-1.0 and labelled as such.
URL: https://github.com/sakura-sky/tollgate

### Arkhe (Apache-2.0)
An ontology language for AI systems. A neutral, version-controlled language for declaring what exists in a domain, how it connects, and what may be done to it, by whom, with what trace, so AI systems can be grounded in it, constrained by it, and audited against it. YAML surface syntax compiles to a canonical JSON model and a tool-contract IR; emitters are deliberately thin. The specification is a contract, never a runtime.
URL: https://github.com/arkhelang/arkhelang

### sql-guard (Python)
Deterministic policy engine for LLM-generated SQL, multi-dialect via sqlglot. Decides whether a generated statement is permitted to run, the same way every time, before it reaches the warehouse.
URL: https://github.com/sakura-sky/sql-guard

## Publications

### The CTO Playbook for Agentic Systems (2026)
A leadership, architecture, and organisational blueprint for AI-native companies. Covers the engineering functions that need named owners, decision rights and autonomy tiers for agents, the economics of verification, the agentic SDLC, a four-layer control architecture mapped to the EU AI Act, NIST AI RMF, ISO/IEC 42001 and SOC 2, and a six-phase roadmap with a gate check before scale. Version 1.2, September 2026.
URL: https://whitepaper.download/ctoplaybook/

### The Governed Agent Trust Environment - GATE (2026)
The framework paper behind GATE: a reference architecture of controls for governing AI agents.
URL: https://deterministicagents.ai/downloads.html

### The Trustworthy Agentic AI Blueprint (2025)
Design principles, missing primitives, and the human-in-the-loop patterns required to deploy autonomous agents responsibly inside the enterprise.
URL: https://whitepaper.download/trustworthyagenticai/

### The Executive AI Playbook (2025)
Strategy, governance, and execution: a working guide for leadership teams operationalising AI across regulated and high-growth businesses.
URL: https://whitepaper.download/aiplaybook/

### The Autonomous Horizon (2026)
A strategic white paper for the executive team on the operating model, the architecture, and the leadership decisions for the autonomous enterprise.
URL: https://www.sakurasky.com/white-papers/autonomous-horizon/

### Signal-to-Noise (2026)
A practical checklist for evaluating AI security tools, built around the four-stage funnel from finding to actionable vulnerability.
URL: https://sakurasky.com/white-papers/signal-to-noise-checklist/

## Other Writing

Still Dangerous: A Practitioner's Guide to Strength, Composure, and Capability for Executives Over Forty. A book on martial arts (forthcoming 2026).

## Selected Essays and Series

- The Regulatory Stack, Parts 1-6 (May 2026): Where AI regulation meets the runtime, and what boards and engineers have to build to close the gap. https://www.sakurasky.com/tags/series-regulatory-stack/
- The Missing Primitives for Trustworthy AI Agents (from October 2025). https://www.sakurasky.com/tags/series-missing-primitives/
- GATE: The Missing Infrastructure Layer for Agentic AI (April 2026). https://www.sakurasky.com/blog/gate-launch/
- The Mythos Ledger, Parts 1-5 (April 2026). https://www.sakurasky.com/tags/series-mythos-ledger/
- The Autonomous Horizon, Parts 1-6 (April 2026). https://www.sakurasky.com/tags/series-autonomous-horizon/
- Engineering the Trustworthy Autonomous Enterprise (December 2025). https://www.sakurasky.com/blog/engineering-the-trustworthy-autonomous-enterprise/
- Your Most Powerful User Is Your Growing Security Blind Spot (August 2025). https://www.sakurasky.com/blog/your-most-powerful-user-is-your-growing-security-blind-spot/
- GCP Cost Reviews: A Framework To Work With Your FinOps Tooling (January 2025). https://www.sakurasky.com/blog/gcp-finops-framework/
- Embracing Zero Trust for Container Security (May 2023). https://www.sakurasky.com/blog/embracing-zero-trust-container-security/
- EU AI Act: Implications for Data and Cloud in Banking, Ecommerce, and Gaming (November 2024). https://www.sakurasky.com/blog/eu-ai-act/
- Reflections on the 2024 DORA Report (December 2024). https://www.sakurasky.com/blog/dora-report-2024-reflections/

Full archive: https://www.sakurasky.com/author/andrew-stevens/

## Experience

### Board Member - Distinct Star (2026 - Present)
A permanent-capital partnership he helped establish, acquiring and backing high-quality cloud, data, security, and AI services firms while founders stay in the business. Long-term strategy, diligence on prospective member companies, and operating perspective at board level.

### Co-Founder - Sakura Sky (2013 - Present)
Cloud, data, and AI consultancy working across North America, EMEA, and APAC.

### Chief Technology Officer - Orbx (2024 - 2026)
Modernised commerce and platform systems across simulation, ecommerce, and digital content. Cloud-native architecture, DevOps and SRE maturity, and the foundations for AI-first systems.

### Co-Founder / CISO - HereToday (2020 - 2025)
Blockchain venture for digital legacy and asset succession. Zero-trust architecture, ML-assisted threat modelling, and compliance readiness across GDPR, CCPA, ISO 27001, and emerging digital asset regulation.

### Head of Data Security - Canopy (2015 - 2016)
Cloud security and data governance controls for a regulated wealthtech platform in Singapore. Reported to executive and board stakeholders on risk, resilience, and security strategy.

### Co-Founder - Jiku.io (2014 - 2016)
Cross-platform content and commerce for publishers and retailers, including Channel Nine and Pacific Magazines.

### Chief Information Officer - APC Global Systems (2014)
Digital transformation across global mining operations: telemetry, edge computing, early ML and computer vision, and IT/OT integration.

### Chief Mobile Officer - HotelsCombined (2012 - 2013)
Owned the mobile roadmap, product, and growth across iOS and Android for a global hotel comparison platform during its expansion across APAC.

### CTO / Head of Product - CarAdvice (2011 - 2012)
Product and technology strategy for one of Australia's largest automotive media properties, aligning editorial, commercial, and engineering behind a single roadmap.

## Technical Depth

- AI governance: NIST AI RMF, ISO/IEC 42001, EU AI Act, OWASP AISVS, MITRE ATLAS, NIST SSDF
- Agent architecture: agent identity and authority, policy-as-code (OPA/Rego), runtime guardrails, tamper-evident audit, observability, kill switches, LLM gateways and spend control
- Cloud architecture: AWS, GCP, Azure, multi-cloud, hybrid, sovereign cloud
- Security: zero trust, cloud security posture, AI security, regulated-industry compliance (GDPR, CCPA, ISO 27001)
- Data and platform: data platforms, MLOps, DevOps and SRE

## Contact

For technology leadership and transformation roles, AI governance, cloud architecture, and security engagements, board and advisory roles, or speaking: andrew@andrewjstevens.com