How the skill is being demonstrated here
Planned → Studying → Practicing → Applied. Applied requires a linked project or build and documentation of how the skill was used.
I'm training for AI GTM Engineering in public. This page separates what I already do in real GTM work, what I'm actively building with, and what I'm formally studying next, so nobody has to guess which is which.
I prove these skills by applying them to real businesses I own and operate. CAROS serves as one of my primary technical environments, while other brands provide additional real-world use cases across different industries, customers, operating models, and business problems.
Planned → Studying → Practicing → Applied. Applied requires a linked project or build and documentation of how the skill was used.
Developing → Working Knowledge → Proficient → Advanced. Professional expertise remains visible when it is practiced in a new technical context.
Studied ≠ Practiced ≠ Applied ≠ Proven. Completing coursework alone does not make a skill Applied.
Using a technology in real work doesn't mean claiming expert-level proficiency. Application status and proficiency remain separate, and the progression stays explicit.
Concepts like RAG, MCP, agent architecture, retrieval, tool use, APIs, multi-model workflows, and AI infrastructure get introduced early because I'm already applying them in real production work. At this stage they're labeled Learning or Practicing, never Proficient.
The same topics come back at greater depth: architecture, certification, production deployment, advanced implementation, security, scalability, governance, observability, performance, and cost optimization.
Where this is heading: Emerging GTM Engineer → AI GTM Engineer → AI Systems / GTM Systems Architect. The destination is designing systems where business process + data + crm + apis + automation + ai models + agents + retrieval + analytics + human judgment work together.
Open any row for the objective, the certification path, related builds, and evidence of real application.
Learn AI capabilities and system architecture, not just AI brands. Different models and platforms can do different jobs inside the same system. The skill is choosing each one for the task, then architecting the whole system around it.
A reasoning and conversational layer.
Development, long-context analysis, coding, and document-heavy work.
Enterprise infrastructure: retrieval, agent deployment, memory, model access, permissions, observability, document processing, and background workloads. Evolved from Vertex AI.
Introduced when they are the best fit for the system being built.
Learning objective: Design and implement AI systems in which models, agents, business data, retrieval systems, APIs, automations, and human workflows operate together securely and reliably.
Loading progress…
Each Learning skill has a target window for its certification. Phase 4 is an intentional market calibration point, not simply another learning phase. The demonstrated portfolio will be evaluated against 20 current, live AI GTM Engineer or closely related job descriptions. Skills that appear repeatedly in the market but are missing from the portfolio will inform Phase 5 and later curriculum.
Phase 1 starts October 15, 2026.
Certs open doors. Builds walk through them. For every AI build I log the model, the cost per run, and the failure rate, and the misses get published right next to the wins.
Oct 15, 2026 to Jan 14, 2027
A Clay table that scores service businesses against the CAROS ICP, with fit signals tuned for each of the 23 industries CAROS serves.
Industry-specific research prompts that hunt for revenue leak signals: missed calls, slow response, review velocity, and front desk hiring. Model, cost per run, and failure rate logged.
Submissions from the Revenue Leakage Calculator and Free Business Audit post straight into the CRM through REST APIs, with every request documented in Postman.
Jan 15 to Apr 14, 2027
Every interaction in the CAROS demo fires a webhook into the CRM, updates contact and deal fields, and moves the opportunity through an 8-stage pipeline. Started early; the webhook and fields are live.
A buying signal triggers enrichment, creates the CRM contact, and alerts the assigned rep with the context needed to open the call.
The CAROS pipeline rebuilt in HubSpot, with dedupe and data sync rules, to prove the architecture travels beyond one platform.
SQL that measures conversion at every step: marketing, leads, calls, bookings, estimates, jobs, customers, repeat business, and revenue.
Apr 15 to Jul 14, 2027
The CAROS pipeline mirrored in Salesforce, plus an agent that drafts next actions on stalled deals.
The same production workflow built twice. Cost, run time, and failure rate compared, and the loser gets retired.
The jobs that outgrew no-code: bulk data cleanup, custom scoring, and scheduled client reports. At least one runs in production as a scheduled serverless function on AWS Lambda or Azure Functions.
Jul 15 to Oct 14, 2027
Before and after metrics from live clients across multiple industries, written up with the systems that produced them.
Every build so far documented with a README, an architecture diagram, and a short walkthrough video.
An intentional market calibration point: the portfolio is evaluated against 20 current, live AI GTM Engineer or closely related job descriptions. Skills that appear repeatedly but are missing from the demonstrated portfolio inform Phase 5 and later curriculum, so what comes next can adapt to the market.
Oct 15, 2027 to Apr 14, 2028
Every marketing dollar traced to booked revenue, built in Power BI and rebuilt in Looker.
Retrieval over CAROS implementation and product documentation so internal answers cite the exact source.
An MCP server that lets AI agents query pipeline and client data safely: secrets kept out of code, OAuth scopes limited to what each agent needs, personal data masked, and HIPAA and FINRA guardrails for regulated industries.
Apr 15 to Oct 14, 2028
A multistep agent that runs the Free Business Audit logic on a real business and returns the three biggest leaks with evidence.
The capstone: enrichment to CRM to automation to BI to agents, documented end to end with two years of results.
A GitHub portfolio with pinned repos and a six-part LinkedIn series covering the full run, misses included.
I am not learning these technologies through hypothetical exercises alone.
I use companies and brands that I own or operate as live environments for experimentation, implementation, and proof of work.
CAROS is one of my primary technical laboratories, but it is not the only one.
Different businesses create different problems to solve. That gives me the opportunity to demonstrate how I choose technologies based on the actual business requirement rather than forcing every tool into the same use case.
The goal is not to say that I studied a technology. The goal is to show what I actually built with it.
Evidence hierarchyBuilds and production evidence remain more prominent than certificates. Certifications validate knowledge. Builds demonstrate capability.
0 published so far
Examples of builds that may become case studies. Not claims of finished work.
Use retrieval, embeddings, vector search, document processing, and tenant-isolated knowledge to support Ask CAROS.
Use multiple models for background classification, enrichment, summarization, and structured processing while keeping Ask CAROS as the primary reasoning interface.
Build an agent that can retrieve business information, use approved tools, perform a multi-step task, log its actions, and escalate appropriately to a human.
Design a workflow where OpenAI, Claude, and Google Cloud AI infrastructure perform different jobs within the same business system.
Combine CRM, attribution, sales activity, communications, and AI analysis to determine what is affecting revenue.
Use one of my other companies or brands to automate a real operating process and document the full implementation.
Build an internal or customer-facing application that combines business logic, data, APIs, AI, and automation.
Designed and shipped carlyfowler.com.
VisitCo-built getcaros.com in English and Spanish.
VisitBuilt an interactive demo prospects can run on their own.
Try itShipped five free revenue tools that qualify visitors before the first call.
Try oneThe first formal cert on the board.
Platform Administrator, earned.
Ten syllabus builds live across the businesses I run.
Two years, documented end to end.
I built this page, carlyfowler.com, the CAROS demo, and all five free revenue tools: strategy, copy, architecture, and code. I co-built the CAROS website and the CAROS dashboard. That’s the point of GTM engineering: understand the revenue problem well enough to design the system—and understand the system well enough to ship it.
How XP works. Proficient skills earn 300 XP, Practicing earns 200, and Learning earns 100. Every certification earned adds 150, and every shipped build adds 100. Each level takes 1,000 XP. Everything on this page is calculated from one data file, so the numbers move the moment a status changes.