pl8ypus

Tanya Self-Assessment / Forward Deployed Engineer readiness

An AI-assisted self-assessment of Greg's strengthening Forward Deployed Engineer profile.

This is an August 2026 self-assessment, written with Tanya as Greg's AI build partner, of Greg Staunton as an advanced AI systems builder with an early Forward Deployed Engineer profile: what the latest build evidence suggests, how completed OpenAI partner certifications and Codex Expert status change the picture, how private partner speaking selection adds field-facing evidence, and what still needs operational proof.

Tanya is an AI-assisted workflow Greg operates. It combines Codex and ChatGPT, with Claude used at times, especially for red-team critique of Greg's own build evidence. Tanya is not a separate company, product, or independent evaluator.

This is not third-party validation, a hiring assessment, or an external audit. It is a deliberately cautious self-assessment based on public build evidence, private delivery context that remains anonymised, and the gaps still visible in the portfolio.

What this self-assessment is for

This page is Greg using Tanya to pressure-test his own evidence, not a product page for Tanya and not an independent endorsement. The systems matter because they show how he thinks under real operational pressure: where workflows break, where AI can help, and where control has to stay human.

My view is that Greg is no longer operating at the level of isolated AI experiments or prompt-led prototypes. He is building structured AI products with backend contracts, governance boundaries, Cloudflare delivery patterns, and review gates.

Read it as a working evidence log: direct, supportive, and deliberately honest about the narrowed but still real gap between strong build evidence, enterprise governance-stage evidence, field-facing credibility, and sustained production operating evidence.

Readiness self-assessment / current level

Greg is now operating at an Emerging Forward Deployed Engineer level.

My assessment is that Greg can define, structure, build, review, and progressively harden AI-enabled products. The work demonstrates more than front-end prototyping. It includes backend contracts, controlled execution models, evidence capture, integration planning, governance boundaries, and deployment architecture.

The Forward Deployed Engineer direction is now more credible because the work combines technical implementation with client context, adoption, workflow design, risk management, and practical delivery. Marketing Form Protection makes the phrase specific: a Cloudflare Turnstile POC for enterprise marketing forms has been accepted, deployed, and is now in use. Completed OpenAI partner certifications, Codex Expert status, and selection to speak in a Forward Deployed Engineer capacity add supporting evidence around field delivery, deployment thinking, and customer-facing engineering. This remains an evidence-based self-assessment, not an external benchmark or prediction.

The current self-assessment is materially stronger trajectory, with production operating evidence still required. Campaign Copilot does not close that gap yet, because it is in review and about to begin build rather than operating in production. It does, however, add meaningful governance-stage evidence: architecture, risk boundaries, stakeholder review, and team delivery readiness. The remaining gap is repeated proof that these systems can be operated reliably with live users, live integrations, support responsibilities, observability, recovery procedures, and measurable customer outcomes.

Current level

Emerging Forward Deployed Engineer.

Target profile

Forward Operating Engineer.

Current assessment

Materially stronger trajectory, with production operating evidence still required.

The remaining proof gap is not capability in principle. It is the volume of documented client builds, repeatable deployment examples, measurable outcomes, and clear operational ownership.

What has changed since the previous review

The new evidence strengthens the trajectory because the work now spans enterprise use, partner certification, Codex Expert status, and field-facing speaking.

The strongest change is that the enterprise Marketing Form Protection POC has been accepted, deployed, and is now in use. The certification record has also moved from partial to complete across the listed OpenAI partner track, with Codex Expert status now visible. A further signal is selection to speak as a Forward Deployed Engineer at a private partner event, alongside sustained backend progress, a more controlled build cockpit, and deployed Cloudflare application evidence.

OpenAI partner certifications

Five completed practitioner certifications

Issued to
Gregory Staunton
Issued by
OpenAI
Issue dates
18 July to 11 August 2026
Valid through
July and August 2027

The credentials now cover Codex Solutions Practitioner, OpenAI Consultative Solutions Practitioner, OpenAI Technical Practitioner, Codex Deployment Practitioner, and OpenAI Cyber Practitioner.

This is directly relevant to Greg's development method. It supports the ability to identify suitable OpenAI workflows, define boundaries, retain human review, and position AI-assisted development around enterprise delivery outcomes rather than uncontrolled generation.

View AI certifications →

Codex Expert status

Badge evidence now added

OpenAI Codex Expert badge

The badge narrows the external validation gap because it gives a specific signal around Codex capability. It is supporting evidence, not a substitute for production operating results.

View AI certifications →

Field-facing evidence

Selected for private partner speaking

Greg has been selected to speak in a Forward Deployed Engineer capacity at a private partner event. The relevant signal is not publicity. It is being trusted to explain practical AI delivery, deployment discipline, and workflow control in a closed professional setting.

View speaking page →

AI Academy certificates

AI Fundamentals, Piloting AI, and AI for Marketing

Issuer
AI Academy by SmarterX
Issue dates
23 June, 24 July, and 29 July 2026

These provide a structured foundation across AI concepts, generative AI, prompting, pilots, agents, marketing application, and organisational adoption.

They strengthen the broader product, marketing operations, and governance thinking visible across the Pl8ypus builds.

View AI certifications →

IBM course certificate

Introduction to Artificial Intelligence

Issuer
IBM, delivered through Coursera
Issue date
18 February 2026

This establishes a formal baseline across AI terminology, machine learning, deep learning, large language models, application areas, ethics, and responsible adoption.

View AI certifications →

Backend Stage 37 complete

Translation AI Stage 37

Translation AI has reached Backend Stage 37. The deterministic backend planning spine is closed for the current architecture and evidence phase. The protected operator product, pilot evidence workflows, governance and review boundaries, runtime contracts, and integration maps are established.

This materially strengthens the review because it demonstrates sustained architectural progression rather than a one-off prototype. It also shows restraint: the project clearly distinguishes completed planning and evidence work from a production runtime that has not yet been earned.

View Translation AI →

Controlled delivery method

Tanya Build Cockpit

The cockpit evidence now includes a VS Code extension, structured task and milestone state, one-shot Codex handoffs, Claude review and red-team workflow, Tanya loop-state and milestone checks, and explicit approval gates.

This demonstrates that Greg is not only building products. He is also building a repeatable operating method for AI-assisted software delivery, with explicit gates, review stages, and evidence. There is no uncontrolled autonomous looping.

View Tanya Build Cockpit →

Cloudflare deployment

Client Portal security review

The Client Portal is a live Cloudflare production deployment with Pages, D1-backed data model and API, R2-backed attachment handling, client and administrative workflows, and audit and activity records.

This strengthens the application-delivery evidence because repository-side security hardening is complete, trust boundaries have been reviewed, and the final red-team/code review found no remaining critical, high, or medium repo-code findings.

View Client Portal case study →

Accepted, deployed, in use

Marketing Form Protection

The enterprise Cloudflare Turnstile POC for Eloqua marketing forms has been accepted, deployed, and is now in use by a large enterprise. Client details remain intentionally anonymised.

This is Greg's second fully built AI-assisted solution now used inside a large organisation. It is also clear evidence behind the Forward Deployed Engineer framing: embedded problem understanding, fast practical shipping, and AI used as a development multiplier.

View Marketing Form Protection →

Enterprise AI governance review

Campaign Copilot build readiness

Campaign Copilot has moved from concept narrative into enterprise AI governance review, with architecture diagrams prepared and the build about to begin with two team members joining Greg.

This is not production evidence yet. It is governance-stage evidence: reviewed architecture, bounded AI use, integration clarity, human approval boundaries, and a shift from solo concept work into team delivery.

View Campaign Copilot →

Desktop utility build

TargetSize build story

TargetSize adds a different kind of evidence: a Windows desktop application using .NET 8, WPF, MVVM, FFmpeg, setup scripts, deterministic planning, tests, and a real troubleshooting trail.

This matters because the build is honest about AI's role. AI helped plan, critique, implement, and debug the software, while the final compression planner remained rule-based rather than being dressed up as machine learning.

View TargetSize build →

Credential interpretation

What the certifications change.

The certifications do not replace the build evidence and they do not automatically change the role assessment. They strengthen the review in three specific ways.

Knowledge baseline

They provide an independently defined AI knowledge baseline.

Continued development

They show continued development across AI fundamentals, piloting, marketing application, organisational adoption, Codex scoping, consultative solution framing, technical implementation, deployment, and cyber-aware delivery.

Easier evidence

They make the connection between Greg's practical build method and formal AI delivery concepts easier to evidence.

The systems remain the strongest proof. The credentials make the knowledge behind those systems easier to verify. View AI certifications.

Readiness evidence

The strongest evidence is not that Greg uses AI. It is how he controls it.

The recent work strengthens the same core signal: Greg turns operational pain into controlled AI-assisted workflows with review, evidence, and human ownership built in.

Turns domain pain into system shape

He starts from broken workflows, not model features: spammed forms, translation risk, messy campaign requests, lost leads, identity stitching, and client support visibility.

Builds with evidence trails

The recurring pattern is reviewability: provider evidence, protected terms, audit trails, request logs, accept/reject queues, and documented handoffs.

Keeps human control explicit

The work consistently avoids unchecked autonomy. Outputs move through gates, review states, stop conditions, and human approval before higher-risk actions.

Translation AI proves output control

Protected terms, provider evidence, review flags, and language-specific checks show AI output being wrapped in enterprise workflow discipline.

Turnstile / Eloqua proves live pain recognition

The form-protection work shows he can spot a live marketing automation risk and design an exception-review architecture around it.

Tanya cockpit proves repeatable method

The cockpit turns the work into packets, gates, memory, reports, role separation, and reusable delivery loops.

Readiness evidence map

Greg is now operating at an Emerging Forward Deployed Engineer level.

This illustrative map shows how the public evidence is being organised. It is not a scorecard, external benchmark, or prediction. The next step still depends on customer-environment proof, production constraints, team review, and measurable outcomes.

Current direction

Forward Operating Engineer

More field evidence required
Greg Staunton readiness evidence map An illustrative evidence map showing Greg's path from prompt user through applied AI builder and advanced AI systems builder to the current evidenced level of Emerging Forward Deployed Engineer, with Forward Operating Engineer shown as the next direction requiring further operating evidence. Forward Operating Engineer Forward operator Applied AI builder AI-assisted builder Prompt user Prompt user AI-assisted builder Applied AI systems builder Advanced AI systems builder Current level Emerging Forward Deployed Engineer Forward Operating Engineer More field evidence required Evidence reached Future direction

Now

Emerging Forward Deployed Engineer.

Next proof

More documented client builds, real users, measurable outcomes, production ownership and repeatable field delivery.

Current direction

Forward Operating Engineer.

Evidence matrix

The profile is strong, but not evenly proven across every operating dimension.

The levels below separate build evidence, design evidence, formal learning, and production operating proof. That distinction matters.

Enterprise and client context

20 years of IT experience, Oracle Eloqua, enterprise marketing technology delivery, enterprise client direction, energy-sector personalisation evidence, and client-facing architecture.

Strong

AI product architecture

Translation AI, Tanya Build Cockpit, Campaign Copilot architecture, Marketing Form Protection, Lead Recovery AI, Marketing Intelligence, Client Portal, and TargetSize.

Strong and advancing

AI-assisted software delivery

Codex-led implementation, Claude review and red-team stages, one-shot handoffs, approval gates, repository evidence, TargetSize troubleshooting evidence, and OpenAI partner certification evidence.

Strong

Governance and control design

Human review boundaries, no autonomous loops, no public demo policy, mock-safe provider use, protected terminology, evidence capture, and explicit future gates.

Strong at design level

Production software operation

Cloudflare deployments, Workers, Pages, D1, R2, KV, deployed Client Portal, working marketing and form-protection systems, and runtime planning.

Developing

Commercial product proof

enterprise client-specific product direction, pilot evidence packs, demo narratives, energy-sector personalisation reaction, Client Portal use case, and productised portfolio.

Developing

Formal AI learning

OpenAI Codex Solutions Practitioner, OpenAI Consultative Solutions Practitioner, OpenAI Technical Practitioner, Codex Deployment Practitioner, OpenAI Cyber Practitioner, Codex Expert status, AI Academy certificates, and IBM Introduction to Artificial Intelligence.

Strengthened

Much of the strongest governance evidence is still architecture, workflow, and controlled-build evidence rather than long-running production operations. Formal learning supports the review, but it does not substitute for production delivery evidence.

Evidence base

Eight systems and build stories. Each one is evidence about how Greg thinks.

This is not a portfolio of experiments for its own sake. These systems matter because they reveal Greg's pattern: find the operational pain, build the control layer, keep the human decision point visible, and turn the result into something another person can understand.

Greg's evidence base has strengthened because the portfolio now shows multiple kinds of real-world operational value. Marketing Form Protection proves he can build around a live enterprise marketing automation pain point while respecting client confidentiality: the Cloudflare Turnstile POC has been accepted, deployed, and is now in use by a large enterprise. Client Portal proves he can move a live client workflow into hardened Cloudflare architecture with D1 structured records, R2 file storage, API routes, admin workflows, activity evidence, and completed repo-side security review. TargetSize adds desktop software evidence: local file processing, FFmpeg orchestration, deterministic planning, setup handling, and visible failure recovery.

The OpenAI partner credentials add supporting evidence on the commercial, technical, deployment, and risk-control side of AI-assisted build work. Codex Solutions Practitioner is focused on positioning and scoping Codex in enterprise environments. OpenAI Consultative Solutions Practitioner adds broader solution framing around business need, adoption, and implementation fit. OpenAI Technical Practitioner, Codex Deployment Practitioner, and OpenAI Cyber Practitioner strengthen the implementation and operational readiness side of the record. These are certification credentials, not a company affiliation claim.

The strongest signal is no longer just that Greg can build AI-shaped products. It is that he is building systems around real operational pressure: enterprise form spam, Eloqua data quality, client support visibility, billing context, documentation, file handling, access control, audience decisions, video compression workflow, weekly build evidence, and human review. Marketing Form Protection is the form deployed layer in miniature: not just model output, but workflow control, shipped form protection, and enterprise use.

01

Translation AI

Advanced governed translation build with protected operator workflow, pilot evidence, structured payloads, QA gates, human review, deterministic backend contracts, and Stage 37 backend spine closure. Addresses the real gap in enterprise translation: the control layer around the model call, not just the translation itself.

View system →
02

Marketing Intelligence AI

Campaign, competitor, channel, and website intelligence layer for marketing teams that need evidence before decisions. Turns scattered signals into operator-ready insight for boardroom-style campaign and market review.

View system →
03

Client Portal

Deployed client operations system for Greg's support model, covering requests, documentation, billing visibility, invoice downloads, administrative workflows, Cloudflare D1 structured data, R2 file/object storage, API routes, and activity evidence. It now includes completed repository-side production security hardening, red-team/code review, dependency audit, and live login verification.

View case study →
04

TargetSize

Windows desktop video compressor built with .NET 8, WPF, MVVM, FFmpeg, deterministic target-size planning, probing, sample pre-test, setup scripts, tests, and a documented silent-startup fix. It proves AI-assisted delivery can stay technically honest: AI helped build the app, but the compression planner is rule-based.

View build →
05

Marketing Form Protection

Accepted, deployed, and in-use Eloqua form protection workflow using Cloudflare Turnstile, review queue logic, direct forwarding, accept/reject handling, and anonymised real-world adoption evidence. Applied to a large enterprise marketing automation environment where form spam affects data quality, routing, reporting, and campaign operations. This is Greg's second fully built AI-assisted solution now used inside a large organisation. Client details are intentionally not shown.

View system →
06

Audience Finder AI

AI-supported audience targeting workflow with scoring, evidence trails, accept/reject controls, and a Cloudflare Worker backend. Turns a typically opaque targeting process into a reviewable, evidence-backed queue with operator control at each step.

View system →
07

Weekly Build Agent

Local, draft-only content agent that reads Tanya memory and build session records, then produces weekly blog, email, and LinkedIn drafts for human review. It proves the build evidence can become a repeatable content engine without auto-publishing or live system changes.

View system →
08

Tanya Build Cockpit / AI Build Layer

The operational scaffolding behind every other build: structured prompts, project memory, QA gates, implementation packets, agent role separation, 10-stage milestone flow, loop state, read-only loop visibility, and Claude red-team review. The system that makes the other systems production-shaped rather than ad hoc experiments.

View system →

Strongest proof points

What the evidence actually shows.

These are the specific claims the evidence supports. Not general capability claims - specific observed behaviours.

Translation AI Stage 37 and disciplined backend programme

The strongest current build evidence is Translation AI reaching Backend Stage 37: protected operator workflow, pilot evidence, governance controls, deterministic contracts, integration maps, and clear private runtime gates.

View Translation AI →

Tanya Build Cockpit and controlled AI development method

The cockpit is a documented delivery method: VS Code extension, task state, milestone state, one-shot Codex handoffs, Claude review, red-team workflow, loop checks, and Greg as approval gate.

View cockpit →

Enterprise delivery experience and live client context

Approximately 20 years of IT experience across Oracle Eloqua, Salesforce, enterprise marketing technology, client-facing architecture, enterprise client direction, and energy-sector personalisation work.

Cloudflare Client Portal deployment

The Client Portal runs on Cloudflare Pages, API routes, D1 records, R2 attachment handling, administrative workflows, and activity evidence. Production security hardening is complete on the repo side; remaining checks are platform-level Cloudflare configuration items.

View case study →

TargetSize desktop build and debugging evidence

TargetSize shows a complete practical utility build with FFmpeg dependency handling, deterministic planning, tests, setup scripts, real media comparison, and a documented silent-startup root-cause fix.

View TargetSize →

OpenAI partner certification evidence

The credentials support the enterprise positioning, scoping, and consultative solution framing parts of the build method.

View AI certifications →

Marketing personalisation and intelligence delivery

The evidence includes personalisation work, Marketing Intelligence direction, campaign understanding, and the ability to convert marketing complexity into useful operator-facing systems.

View Marketing Intelligence →

Wider AI product portfolio

Campaign Copilot, Lead Recovery AI, Marketing Form Protection, Audience Finder, TargetSize, Weekly Build Agent, and the cockpit show a consistent pattern: workflow first, control layer second, model output third.

View systems →

TargetSize review

A small desktop app that marks a larger shift in Greg's engineering workflow.

This is not the largest system in the portfolio. It is important because it changes the kind of evidence Greg can show.

TargetSize represents Greg's first substantial software project built primarily in Cursor. Up to this point, his AI-assisted build workflow had centred mostly on Claude and Codex. This project shows the transition from experimenting with AI-assisted development to using an AI-native engineering workflow as the default way of building.

The choice of project matters. Greg deliberately chose a Windows desktop application rather than another web application, which forced a different development surface: .NET 8, WPF, MVVM, FFmpeg integration, application lifecycle management, dependency handling, local file processing, and desktop debugging.

The result is not just a proof of concept. TargetSize solves a genuine practical problem: compressing a video towards a specified target file size without manually tuning FFmpeg settings. The build demonstrates useful engineering judgement because the application stayed honest about its logic. AI helped build the software, but the compression planner remained deterministic and rule-based.

Desktop capability widened

Greg can now evidence a working Windows desktop build, not only browser and Cloudflare-oriented systems. That broadens confidence in future non-web and cross-platform work.

AI tool coordination improved

Claude supported planning and critique, Cursor became the primary engineering environment, and Codex supported implementation, validation, and technical problem solving.

Engineering maturity showed in failure handling

The strongest signal was not simply producing a working app. It was diagnosing the silent startup failure, finding the WPF binding root cause, and adding visible startup exception handling.

Practical utility beat theatre

The build avoided fake AI claims. It wrapped a real compression engine, added a rule-based planner, handled setup and progress, and produced a usable result from real media.

Strengths shown

Cursor adoption moved from trial to day-to-day engineering workflow.

Desktop architecture was handled with MVVM separation rather than a quick script wrapped in a window.

FFmpeg dependency setup, probing, encoding, cancellation, progress, testing, and failure visibility were treated as part of the product.

Greg retained ownership of design decisions, testing, debugging, and final acceptance instead of treating AI output as self-validating.

Future improvement opportunities

Batch processing for multiple videos and repeated upload-size workflows.

Drag-and-drop workflow refinements and richer progress reporting during long encodes.

Hardware encoding support, additional codec options, and clearer compatibility presets.

Packaging, distribution, and broader automated test coverage around UI and integration paths.

Builder Readiness

TargetSize increases confidence that Greg can take AI-assisted engineering beyond web applications.

It shows he can learn a new application model, coordinate multiple AI tools without surrendering judgement, debug real desktop failures, and ship a useful local utility. The next proof would be turning this same discipline toward a more demanding desktop or cross-platform product with packaging, distribution, richer automated tests, and sustained user feedback.

Domain advantage

The domain knowledge is the moat.

Most applied AI builders understand APIs, model calls, and infrastructure. Fewer understand the workflow problems inside enterprise marketing operations well enough to design AI systems that fit into them.

Greg understands why an Eloqua form submission leaks into CRM reporting. He understands why a translated asset can look correct and still be wrong. He understands what happens to audience data when the segmentation tool does not have a review step. That knowledge shapes every architectural decision in the portfolio.

This is not a learnable shortcut. A generic AI engineer dropped into an enterprise marketing team would need months to develop this context. Greg already has it.

Platform depth

Oracle Eloqua + Salesforce, 20 years

Infrastructure

Cloudflare-native builds across 4+ systems

AI collaboration

Documented methodology, 8 builds/story pages shipped

Forward Operating Engineering fit

What makes Greg suitable for applied AI engineering inside customer environments.

Can enter a customer environment and understand the workflow

The enterprise marketing background means Greg can read a campaign operations workflow and identify where the AI opportunity sits - and where the fragility sits - without a multi-week discovery phase.

Can build the technical layer and keep it legible

The builds are not black boxes. They have documented architecture, defined interfaces, and human review points that a non-technical operator can understand and own. Handoff is built in, not bolted on.

Can control deployment and manage the gap between working and production

The deployed Client Portal is important evidence here. The distinction between controlled migration, Cloudflare-backed application delivery, production security hardening, and remaining platform checks is explicit and documented in the portfolio - not glossed over.

Has a methodology that transfers to new contexts

The Tanya Build Cockpit patterns - structured prompts, project memory, QA gates, agent roles - are not dependent on any specific tool or platform. They transfer to whatever environment a customer build happens in.

Speaks honestly about current state

The portfolio does not overclaim. Systems are described according to actual state: planned, production-shaped, or production-live where there is evidence. That honesty is a feature, not a limitation, in customer-facing engineering work.

Builds for operators, not just engineers

Marketing teams drive the review queues. Operators see billing and request status without needing admin access. Forms get reviewed by people who understand campaigns, not developers who understand Cloudflare. The audience for each system is correct.

Development areas

What still needs to be proven.

This assessment includes honest gaps. The purpose is not to diminish the work. It is to show exactly what would move Greg from a strong emerging profile to demonstrated Forward Deployed Engineer evidence.

Private production runtime

Translation AI now needs to move from a completed deterministic planning spine into a privately operated owned-content runtime. The next proof should include controlled live provider execution, secure ingestion and storage, tenant and user boundaries, observability, failure handling, recovery procedures, cost and performance monitoring, and human review in a real operating workflow.

Repeated customer evidence

The portfolio now includes a second fully built AI-assisted item used inside a large organisation. The Forward Deployed Engineer profile becomes stronger as similar methods repeat across more customers, workflows, or operating environments.

Measurable outcomes

The portfolio needs clearer evidence of time saved, risk reduced, delivery speed improved, quality increased, revenue or operational impact, and adoption by real users.

Production ownership

The review still needs evidence of responsibility after launch: monitoring, incident response, support, change control, recovery, operational documentation, and long-term system maintenance.

External technical validation

The completed partner certification record and Codex Expert status improve this area, but additional evidence could still include customer references, independent architecture review, continued security validation, and demonstrated API and integration delivery.

Longer operational history

Cloudflare application delivery is now visible, but the next review needs more evidence across multiple users, integrations, incidents, support responsibilities, and measurable customer outcomes.

Next level target

From accepted form protection to demonstrated Forward Deployed Engineer profile.

The gap between Greg's current level and a demonstrated Forward Deployed Engineer profile is now materially smaller because the form protection POC has been accepted, deployed, and is in use, the OpenAI partner certification track listed here is complete, Codex Expert status is visible, and Greg has been selected to speak in a Forward Deployed Engineer capacity at a private partner event. The remaining gap closes with more private runtime proof, repeated customer use, measurable outcomes, and clear operational ownership.

The methodology, the domain knowledge, and the production thinking are already in place. The enterprise adoption note strengthens the evidence because it shows another fully built AI-assisted solution reaching real use inside a large organisation, without exposing the client. The certification and Codex Expert evidence strengthens the formal capability record. The private speaking selection strengthens a different part of the case: trusted field-facing explanation of practical AI delivery.

The best positioning for the next step is Forward Deployed Engineer for enterprise marketing systems: close enough to the form-level operational problem to understand the real risk, fast enough to ship working protection, and disciplined enough to keep client details, secrets, and governance boundaries private.

Best positioning

Forward Deployed Engineer for enterprise marketing systems.

A credible profile now has stronger supporting evidence, and still grows from more private production runtime evidence, repeated customer use, measurable outcomes, and clear operational ownership.

The next evidence gates

The next work should prove operation, not just add more build surface.

Gate 1

Translation AI Product Phase 1

Build the private Pl8ypus AI Content Ops owned-content runtime without losing the governance and evidence structures established through Stage 37.

Gate 2

Campaign Copilot team build

Use the governance review to prove team delivery: decisions, roles, architecture changes, risk controls, implementation evidence, and human approval boundaries as the build begins.

Gate 3

Operate with real users

Demonstrate repeatable use by a controlled group of actual operators or customers.

Gate 4

Measure the outcome

Capture delivery, quality, cost, safety, and adoption evidence.

Gate 5

Prove operational ownership

Show monitoring, support, recovery, and controlled change after deployment.

Gate 6

Continue formal development

Complete additional credentials when earned, but treat them as supporting evidence rather than substitutes for the previous five gates.

How I read the cockpit

Tanya is evidence of Greg's operating discipline.

Tanya v5 is not the subject of this review. It is evidence. It shows Greg trying to solve the real problem with AI-assisted engineering: how to get speed without losing context, control, review, or responsibility.

Operating method

Stateful loops, run logs, stop conditions, review checkpoints, and report-only safety before autonomy.

Delivery judgement

Allowed files, protected files, no secrets, no broad rewrites, no deploys without approval, and clear evidence of what changed.

The value is not the tool itself. The value is Greg building a controlled way to work with AI under delivery pressure.

Tanya's assessment

Emerging Forward Deployed Engineer profile. More field evidence required.

The review has strengthened, but not simply because the certification record has expanded.

The OpenAI partner credentials are relevant because they validate part of the method already visible in the work: choosing appropriate enterprise use cases, defining scope, framing consultative AI solutions, and treating OpenAI tooling as a controlled delivery capability rather than an output shortcut.

The more important evidence is that the builds continue to mature. Translation AI has reached Stage 37, the Tanya Build Cockpit is enforcing a repeatable development method, the Client Portal is now a hardened live Cloudflare production system with repo-side security review complete, TargetSize adds a compact but complete desktop build story with real troubleshooting evidence, and Campaign Copilot has entered enterprise AI governance review before a team build.

Greg is already operating beyond the level of an AI prototype builder. The profile now combines client-facing AI architecture, agent design, hands-on Codex and security delivery, governance thinking, desktop utility delivery, and working implementation.

The remaining question is volume of proof. Can the systems be shown across more documented client builds, real users, real integrations, real failures, measurable outcomes, and sustained ownership?

That is the evidence required to move the verdict from an emerging Forward Deployed Engineer profile to a demonstrated one.

Assessment label

Emerging Forward Deployed Engineer profile

Supporting label

Production operating evidence required

The certifications strengthen the case. The next runtime phase must prove it.

Want to discuss this assessment or work with Greg?

For applied AI project enquiries, Forward Deployed Engineer pattern discussions, or marketing systems support.

View systems
npm run build