Explore · preview
Marketing leadership: career records, anonymized cases and the live tools, each labeled for the evidence behind it.
I declined to renew a PPC vendor whose reporting did not show persuasive downstream evidence and redirected the money into a bounded, quarter-long experiment.
I managed three Senior Managers, each leading a team, with hiring, coaching and performance reviews as part of the role.
A free-trial motion needed a way to tell whether a trial conversion represented real progress: whether it could be connected through provisioning, activation, product usage, qualification, sales follow-up, and revenue.
Read the case study →Score your free trial motion across five layers: offer, activation, lifecycle, routing, and reporting. Get your primary constraint and prioritized next moves.
Open the tool →I rebuilt the enterprise ABM and demand-generation programs and the scoring and lead flow beneath them. SQLs grew 30% and sales pipeline 15% within six months, with no single initiative credited.
Trial pages were being built one-off per campaign, which slowed iteration, drove inconsistent messaging, and made conversion learning difficult to compare across waves.
Read the case study →Translate a revenue target into the pipeline, closed-won deals, opportunities, and leads or PQLs required to support it.
Open the tool →Campaign data existed. Decisions did not. Reporting was descriptive but not actionable, which left interpretation work to the audience.
Read the case study →Turn campaign performance into a structured diagnosis, recommendation, and decision.
Open the tool →Register one experiment, run the pre-flight feasibility check, and see what the evidence actually licenses — adopt, continue, revise, or stop.
Open the tool →Brandon Olander, Demand Generation Systems Operator. Greater Seattle Area.
Contact →Campaigns were launching, but ownership, sequencing, and stakeholder visibility were inconsistent. Cross-functional teams were spending too much time renegotiating how work moved instead of executing the work itself.
Read the case study →Inject a constraint into a demand system and trace what breaks downstream, what stays healthy, and where to intervene.
Open the tool →Trace how product usage, buying-group evidence, customer context, and governance turn fragmented GTM signals into coordinated revenue action.
Open the tool →Campaign documentation, QA, and readouts were consuming time disproportionate to the value produced. Quality drifted whenever execution pressure rose.
Read the case study →Engineering work you can inspect: a browser simulation of control-system concepts and its documented roadmap.
A control-system simulation of an automated quality cell: explicit machine states, interlocks, fault handling, a simulated E-stop, and recovery that never restarts production on its own.
A browser simulation of control-state behavior only: not a safety PLC, safety-rated control or a certified emergency stop, and not evidence of machinery-safety compliance. Its E-stop is a simulated input to a simulated controller, and nothing here replaces a risk assessment or physical machine validation. AI-assisted tools helped build it. Its tests check the simulation against its documented model, not a physical machine. Neither the simulation nor its tests establish independently demonstrated control-engineering competency on my part.
Open the simulation, with its full limitations and AI-assisted development disclosure →Machine behavior here is determined by explicit state, inputs, interlocks, and controller rules — not by what the interface happens to allow you to click.
The interface never sets an actuator. A button posts a command; the controller evaluates that command against the current state and the interlocks, and only then does an output change. That is why an illegal action produces a stated reason instead of silently doing nothing.
A permissive is a condition that must hold before an action may begin — the checklist at the door. An interlock is a rule that holds throughout, whatever anyone asks for. Start has seven permissives; the conveyor has an interlock that keeps it off while the machine is faulted.
Machine state and process phase are separate concepts. The machine state says what the controller is allowed to do. The process phase says where a part is in the cycle. A cell is RUNNING while it moves through waiting, releasing, transporting, inspecting, sorting and completing — those are not machine states.
These change the environment the machine is reacting to, not the rules it follows. That distinction is the point: an operator command asks the controller for something, an injected condition simply becomes true.
Recovery is deliberately unhelpful. Clearing a jam does not restart production. Releasing the simulated E-stop does not restart production. A successful reset always lands in Idle and never in Running, so returning to production is always an explicit operator decision.
See it on the Aegis page →Times are controller ticks measured from power-on, not your clock. The same command sequence always produces the same log.
AI-assisted tools contributed to building this project. What it shows are implementation artifacts: the simulation code, its documented control model and its design records. Automated tests and an independent review check that the simulation behaves as that model specifies; they test a simulation, not a physical machine. Neither the implementation nor its passing tests establishes independently demonstrated control-engineering competency on my part.
See it on the Aegis page →Project Aegis is a multi-milestone build. This page is M0 and M1: the deterministic control foundation — explicit state, interlocks, faults, recovery, and machine telemetry — and nothing beyond it. The name of the page describes what is actually running, not what the project intends to become.
Later milestones extend the same controller: a richer process model, alarm history and operator authority, production quality analysis, industrial cybersecurity, and a physical prototype with the control model translated toward a PLC.
A constrained diagnostic AI layer belongs to a later milestone too, and it is bounded by design: advisory only, never required for safe machine operation, and never in control of a production-critical output. None of it exists yet, and nothing on this page pretends otherwise.
The broader “secure autonomous” description belongs to the finished system once security and autonomy have actually been built and tested. This milestone has not earned it.
See the roadmap on the Aegis page →On this floor: 1 browser simulation and 1 documented roadmap. No hardware build is claimed here.
The engineering lab →Interests, experiments and two self-reported personal projects outside work, each labeled for what it is, and one engineering case study at the software desk. Apart from that case study, no documentation of a personal project is published yet.
Small tools I build for my own use, with AI-assisted development. They are unpublished experiments: no production use is claimed, and they are not evidence of independently demonstrated coding competency.
An applied systems-engineering case study in AI-assisted software development, automation, verification, and controlled autonomy.
Human-directed, AI-assisted: Claude Code agents implemented nearly all of it, and every AI review was a separate Claude session. It is not evidence of software-security credentials, industrial-controls safety competence, or unaided coding.
Read the case study →Printers, materials and printable parts. An interest; no printed project is published here.
Grinders, extraction and the equipment behind a good shot. An interest; no project is claimed.
First-person-view flying and the equipment path into it. An interest; no build is claimed.
Connected devices and home-lab gear. An interest; no build is published here.
Hands-on building and small fabrication. An interest; nothing is published yet.
Cars, electric mobility and the equipment around them. An interest; no build is published here.
I built and customized an electric longboard, including its battery pack and acceleration settings. Self-reported: no documentation of it is published yet, and no performance or safety figure is claimed.
On my Tesla Model 3 I added customized puddle lights, interior LED lighting and a powered frunk. Self-reported: no documentation of the work is published yet.