Jun (Joel) He, PhD
Engineering Governed AI Systems. Building the Organizations That Deliver Them.
Distributed-systems researcher, cloud infrastructure architect, entrepreneur, and engineering leader. My work spans large-scale production infrastructure, engineering organizational design, and the emerging control planes required to operate autonomous AI systems safely and reliably.
Years in Distributed Systems
Systems Publications
Engineers Led
"A sovereign system is not safe because a model behaves correctly. It is safe because no model can cross the execution boundary without deterministic admission."
Research Depth. Hyperscale Execution. Organizational Scale.
Bridging rigorous academic research in distributed optimization with over a decade of tier-0 infrastructure engineering at AWS and entrepreneurial engineering leadership.
Research & Systems Foundations
PhD in distributed systems, networking, and optimization theory from Princeton University. Authored 20+ peer-reviewed papers spanning distributed consensus, network optimization, and queuing dynamics, bridging rigorous theoretical models with production systems reality.
Hyperscale Cloud Infrastructure
More than a decade of engineering experience at Amazon Web Services (AWS), delivering tier-0 cloud control-plane services, high-availability data planes, and automation frameworks where service degradation carried immediate global operational impact.
Entrepreneurship & Engineering Leadership
Former technology founder and engineering leader with experience directing an engineering organization of up to 200 engineers in an earlier company. Experienced in building high-caliber teams from zero, developing managerial talent, and aligning technology execution with business strategy.
Agentic Infrastructure & OpenKedge
Founder and Chief Systems Architect of OpenKedge. Authored Post-Deterministic Distributed Systems (PDDS) and Persistent Cognitive Identity (PCI), establishing the reference control plane that decouples probabilistic AI reasoning from deterministic, policy-bound execution.
Autonomy scales when intent, decision authority, and feedback are explicit.
My leadership approach has been shaped by entrepreneurship, research, and hyperscale cloud engineering. Across these environments, I have learned that sustainable engineering performance depends less on centralized task assignment or activity metrics than on clear priorities, technical ownership, capable teams, and rapid organizational decisions.
Autonomy-Based Engineering Management
Core Principle"Centralize intent and accountability; decentralize implementation choices."
Engineering Organizational Resilience
Core Principle"High-performing organizations invest in trust, continuity, and clarity—not just delivery activity."
Cross-Organizational Infrastructure Leadership
Core Principle"Technical leaders must combine architectural rigor with institutional execution."
From Managing Tasks to Governing Autonomy
AI-assisted development is reducing the cost of implementation, but software systems still depend on sound architecture, effective product decisions, human judgment, validation, and organizational coordination. The next generation of engineering leadership must evolve accordingly.
I believe engineering organizations should operate through clear intent, distributed execution, explicit decision rights, durable commitments, and evidence-driven feedback. These principles reduce coordination overhead while protecting quality and accountability.
This organizational philosophy complements my technical research on governing autonomous AI systems. They are different domains, but both require well-defined boundaries between autonomy and authority.
Customer evidence over activity metrics
Value is defined by observable customer outcomes and production reliability, not vanity metrics like lines of code, PR volume, or token consumption.
Engineering autonomy within clear priorities
Engineers pull high-leverage work autonomously against transparent, stack-ranked organizational priorities rather than waiting for assigned tickets.
Durable decisions and accountable ownership
Decisions are codified in append-only Architectural Decision Records (ADRs) with explicit Type 1 vs. Type 2 authority, eliminating decision paralysis.
Safety and verification by design
Automated admission controls, contract tests, and blast-radius isolation ensure system invariants are guaranteed before changes reach production.
Organizational learning over bureaucracy
Replace multi-tier approval committees with rapid telemetry feedback, blameless operational reflection, and continuous adaptation.
When Code Becomes Cheap, Leadership Becomes the Bottleneck
A practitioner essay examining why AI coding tools increase implementation capacity without improving strategic judgment, and detailing the six-pillar operating model for engineering leaders.
Sovereign Compute Is Complete Only When Execution Is Sovereign
Localized infrastructure secures where models and data reside. Autonomous operations must also govern who may act, under which policy, and with what evidence. Three risks remain beyond the data-center boundary:
Data Residency ≠ Sovereignty
Local weights and data do not guarantee autonomy if models remain subject to external policies, remote updates, or de-platforming. Sovereignty begins with an in-jurisdiction execution control plane.
Cognitive Supply Chain Risk
No external or black-box model should hold standing credentials over critical systems. Every proposed state mutation must pass deterministic local policy gates.
Institutional Amnesia & Drift
Model upgrades can sever compliance trails, memory lineage, and audited behavior. Governed cognitive lineage preserves continuity across every substrate change.
Govern What Acts. Preserve Who Acts.
PDDS governs the boundary between intent and execution. PCI preserves the identity, memory, and authority of the reasoning entity across every model and substrate change.
PDDS — Post-Deterministic Distributed Systems
Governs Action · Neutralizes Stochastic Drift
Treats every model output as an unvalidated proposal. PDDS evaluates proposed operations against local policy engines before granting ephemeral, scoped execution tokens. No model receives standing execution authority.
PCI — Persistent Cognitive Identity
Governs Lineage · Neutralizes Model Vendor Lock-In
Preserves continuous institutional memory, consent states, and fiduciary compliance as models upgrade across generations or migrate between foreign clouds and sovereign execution enclaves.
National-Scale Autonomy Without Operational Exposure
Leverage frontier reasoning from any global or domestic provider without relinquishing control over proprietary data, operational APIs, or physical execution rights.
Air-Gapped Cognitive Boundaries
Permitting frontier reasoning over industrial telemetry without exposing raw sovereign data, credentials, or physical actuators.
Federated Sovereign Topologies
Connecting multi-agency systems under shared assurance while each domain retains independent, immutable roots of authority.
Cognitive Identity & Succession
Preserving organizational memory, policy compliance, and audit lineage across generations of AI models and substrates.
Tamper-Evident Evidence Chronology
Cryptographically sealing every proposed intent, policy validation, and execution result into durable evidence chains.
Design the Sovereign Control Boundary
Engage on sovereign AI reference architectures, PDDS control plane deployment, PCI identity governance, or executive briefings for state enterprises, sovereign wealth funds, and critical infrastructure operators.