Agent 01Code
Build, test, and ship.
Cloud Automation · AI Infrastructure · Agent Systems
Frank Zhang
I build AI infrastructure, automate the cloud, and design agent systems that think, act, and keep running.
Cloud automationRepeatable infrastructure and operational systems.
Agent systemsReasoning, orchestration, memory, and tool use.
AI infrastructureObservable, resilient systems designed to survive real workflows.
Scroll to evolve
02 / EVOLUTION
Infrastructure
became intelligence.
I learned systems from the bottom up — mainframe and infrastructure, cloud engineering, then years of automation. AI agents are not a reset; they are the next layer of the same discipline.
The through-line: make complex systems repeatable, observable, governable, and useful to the people operating them.
2017MainframeRBC · systems support
2018InfrastructureTTC / Canadian Tire
2019CloudTD · cloud hosting
2020 →AutomationGeotab · cloud platform
NowAI SystemsAgents · memory · LLMOps
03 / ORCHESTRATE
Systems that
work together.
My agent work is less about a single clever prompt and more about the whole operating system around it: planning, research, coding, observation, automation, and the judgment that ties them together.
Agent Orchestrationplan · delegate · observe · improve
Agent 02Research
Read, compare, extract signal.
Agent 03Learn
Practice, adapt, improve.
Agent 04Observe
Watch systems and surface drift.
Agent 05Automate
Turn repeat work into flows.
Human layerCuriosity
Ask better questions and choose what matters.
04 / FEATURED PROJECT
Agent
Memory Bridge.
Persistent memory and governance for long-running agent workflows. AMB carries useful state across sessions while controlling what should be recalled, trusted, revised, or suppressed.
persistent memoryversioned writesgovernanceevaluation
Governed
Designed to combine context assembly with explicit recall and governance checks, rather than relying on simple retrieval alone.
Current Contextgoverned recall
Past sessionDecisionwhy this path was chosen
KnowledgeArtifactvalidated output + evidence
RevisionSuperseded statekept but no longer preferred
SignalPreferencerelevant context, not clutter
PolicyGovernanceauthority · lifecycle · safety
05 / SELECTED SYSTEMS
Built to work
outside the demo.
My strongest work sits where infrastructure, automation, and agent behavior meet. The implementation changes, but the standard stays the same: reliable enough to keep using.
Developer systems
Automation & platform engineering
Six-plus years at Geotab moving from cloud automation into senior-level platform work, with GCP, Kubernetes, Python, CI/CD, and developer automation as the foundation.
See experience ↓
Agent infrastructure
Long-running agent workflows
Multi-session agent experiments focused on orchestration, memory, evaluation, tool use, recovery, and keeping useful state without letting stale context take over.
View AMB on GitHub ↗
Research automation
Evidence before confidence
Research agents and evaluation harnesses that combine data pipelines, retrieval, classification, and measurable feedback instead of relying on a convincing-looking answer.
More on GitHub ↗06 / EXPERIENCE
Systems,
over time.
My career has stayed close to infrastructure even as the layer moved upward — from operations and cloud hosting to automation, platforms, and now AI systems.
Dec 2025 — now
Senior Cloud Automation
Geotab
Cloud automation, platform engineering, and increasingly AI infrastructure / agent-oriented developer tooling.
Apr 2020 — Nov 2025
Cloud Automation Engineering
Geotab
Built and operated automation around cloud infrastructure and developer workflows; the long-running systems foundation behind the work I do with agents today.
Jul 2019 — Apr 2020
Cloud Engineering Associate
TD
Cloud hosting and engineering, after earlier infrastructure and operations internships.
2017 — 2018
Infrastructure & systems roles
RBC · TTC · Canadian Tire
Mainframe support, infrastructure procedure and control, cloud operations, automation, and release management.
07 / BUILDING BLOCKS
The tools change.
The system thinking stays.
Cloud infrastructure
GCP, Kubernetes, deployment systems, reliability, and the operational context AI workloads eventually have to live in.
Platforms & runtime
Infrastructure that gives developers and agents repeatable environments, controlled state, and predictable failure behavior.
Automation
Turn manual repetition into observable workflows — with checkpoints, tests, feedback, and a rollback path.
Evaluation & observability
Measure whether the system actually improved: correctness, drift, context quality, latency, cost, and failure modes.
08 / CONNECT
Build something
that keeps running.
Senior cloud automation engineer in the Greater Toronto Area, building toward AI infrastructure, agent systems, and developer platforms.
Emailsend message ↗
GitHubcode & projects ↗
LinkedInexperience ↗
Back to topreturn to top ↑
Frank ZhangGreater Toronto Area · Cloud Automation / AI Infrastructure / Agent Systems