Evidence type
19 nodes · 3 flows · 3 specialized AI tools
← Automation case studies
Automation case study · AI application architecture
One webhook. Three capabilities. AI only where the decision actually needs AI.
The AI Resume Builder is a compact workflow-backed application prototype built around one stable POST /resume contract. n8n owns routing, AI/tool execution, file conversion and delivery so the frontend can remain thin and replaceable.
Proof summary
Read the evidence before reading the outcome.
Evidence status
Workflow-backed application prototype
Measurement / operating scope
Webhook-first routing, specialized AI tools, structured outputs, PDF rendering, email delivery and evaluation
Observed state
The workflow-backed prototype demonstrates the application contract and three bounded capabilities.
Claim boundary
No production adoption, resume-quality lift, job-placement impact, throughput, uptime or commercial outcome is claimed from the prototype.
Proof reviewed
2026-09-25
Direct answer
What does this case actually demonstrate?
It demonstrates a reusable pattern for AI-enabled applications: define the client contract first, route known intent deterministically, isolate AI inside specialized execution steps, constrain model output, and keep credentials plus third-party side effects behind the orchestration boundary.
Application boundary
Client → webhook contract → n8n orchestration → standardized result.
The interface is deliberately decoupled from workflow internals. If another application can send the same HTTP request, it can reuse the same backend capabilities without duplicating the orchestration logic.
01
Thin client
The frontend collects user input and calls a stable HTTP contract instead of owning AI credentials, file conversion or orchestration logic.
02
Deterministic router
body.flow resolves create, download or check before any probabilistic reasoning is introduced.
03
Specialized execution
Each branch owns one operation and invokes only the AI, file or integration capabilities needed for that operation.
04
Standardized output
HTML, PDF delivery and evaluation reports return through predictable application boundaries rather than ad-hoc agent responses.
Three application flows
Different business operations behind the same API surface.
FLOW 01
Create
Input
Candidate profile · target role/JD · tone · template
Process
Resume Builder orchestrator → validate template → Classic / Modern / Creative tool → structured output
Output
Complete HTML resume returned to the application
FLOW 02
Download
Input
Approved resume HTML · user identity/email
Process
HTML → file → Gotenberg HTTP call → PDF render
Output
Generated PDF attached and delivered through Gmail
FLOW 03
Check
Input
Uploaded resume · base64 PDF payload
Process
Base64 → file → extract PDF text → Resume Evaluation agent
Output
Criterion scores · overall score · prioritized recommendations
AI boundary
Probabilistic inside. Deterministic outside.
The case is useful because it shows where AI should stop. Routing, rendering and delivery are controlled operations; generation and evaluation are bounded interpretation tasks.
Known intent
DeterministicThe Switch router handles create, download or check because the request already declares the operation.
Template generation
AI-assistedThe orchestrator delegates content generation to Classic, Modern or Creative specialist tools.
Output shape
ConstrainedStructured output parsing turns model generation into the HTML contract expected by downstream application steps.
PDF rendering
DeterministicGotenberg converts approved HTML to PDF; the model does not control the side effect.
Email delivery
DeterministicGmail delivery occurs in the download branch with credentials kept backend-side.
Resume evaluation
AI-assistedExtracted resume text is interpreted by a dedicated evaluation agent that returns a structured report.
Workflow evidence
What the exported system can substantiate.
Webhook contract
The workflow export contains a POST /resume webhook used as the single application-facing entry point.
Switch routing
The exported workflow branches on body.flow for download, create and check.
Specialized generation tools
Classic, Modern and Creative tool nodes provide distinct resume-generation behavior behind one orchestrator.
Structured output parser
Generated content is constrained before it is returned to the application or passed downstream.
Gotenberg PDF rendering
The download flow converts HTML to a file and calls a backend PDF-rendering service.
Gmail delivery
The rendered PDF is attached and sent from the backend-side workflow.
Resume evaluation path
The check flow converts base64 to a PDF file, extracts text and sends that text to a dedicated evaluation agent.
Technology footprint
n8n is the orchestration boundary, not the user interface.
n8nWebhookGeminiAI ToolsStructured OutputsGotenbergGmail
The stable contract lets the frontend and model provider evolve independently as long as the input/output boundary remains compatible.
Claim boundary
Workflow-backed prototype — not a hiring-outcome claim.
The evidence supports the exported 19-node workflow, one webhook entry point, three routed application flows, three specialized generation tools, PDF conversion, email delivery and resume evaluation architecture. It does not establish production scale, uptime, model accuracy, ATS success, recruiter acceptance, job-placement lift, conversion rate or ROI.
Related D2 capabilities
The reusable asset is the application boundary, not the resume use case itself.
AI-Assisted Workflows
Bound AI interpretation and generation with explicit routing, validation, state and downstream controls.
Explore servicen8n Automation
Use n8n as an orchestration layer across webhooks, AI tools, files, APIs and delivery systems.
Explore serviceAPI & Webhook Integration
Design stable application contracts so frontends can change without rebuilding the operating logic behind them.
Explore serviceFAQ
Questions this application architecture is meant to answer.
What does the AI Resume Builder system do?+
It is a webhook-first application prototype with one POST /resume endpoint and three routed capabilities: create a resume, render and deliver an approved resume as PDF, or evaluate an uploaded resume and return a structured report.
Why does the workflow use deterministic routing before AI?+
The requested operation is already explicit in body.flow, so a Switch router resolves create, download or check without asking a model to infer known intent. AI is introduced only inside the branch where generation or evaluation is actually useful.
How is resume generation structured?+
The create flow passes candidate and job-description context into a Resume Builder orchestrator, validates the requested template, routes to a Classic, Modern or Creative specialist tool, constrains the result to complete HTML and returns it through the webhook contract.
How are PDF delivery and resume checking handled?+
The download flow converts approved HTML into a file, sends it to Gotenberg for PDF rendering and delivers the PDF through Gmail. The check flow converts an uploaded base64 PDF to a file, extracts text and passes that text into a Resume Evaluation agent that returns a structured HTML report.
Is this a production-scale resume platform?+
No. The published evidence supports a workflow-backed application prototype with 19 nodes, three routed flows and three specialized generation tools. It does not establish production traffic, uptime, conversion, hiring outcomes, model accuracy or ROI.
Your application workflow
Need a reusable backend contract instead of wiring every AI feature into the frontend?
D2 can map the request contract, deterministic routes, AI/tool boundaries, file operations, third-party integrations and failure behavior before implementation.