AI Automation
AI Integration
AI CV Screener is an automated recruitment screening workflow developed for MugDevs to reduce the manual effort involved in reviewing job applications. Built with n8n, Gemini, Google Drive, and Google Sheets, the workflow retrieves candidate CVs, extracts readable content from PDF documents, and evaluates applicants against criteria defined for the hiring role. Gemini generates structured candidate assessments, including scores and supporting explanations. The workflow organizes these results in Google Sheets, providing recruiters with a consistent format for reviewing and comparing applicants. Documents that cannot be reliably processed, including unsupported file formats and scanned PDFs without extractable text, are flagged for manual review rather than assigned unreliable scores. MugDevs reported that the initial screening of more than 70 CVs, previously taking approximately one week, was reduced to minutes. This was a paid client engagement completed between April and May 2026. The workflow is currently idle between hiring rounds, with maintenance provided as needed.
Project details
Context
MugDevs needed a more efficient way to perform the initial screening of job applications. Reviewing CVs manually required recruiters to open individual documents, examine candidates' qualifications and experience, compare those details against job requirements, and organize the results for further consideration. This process became particularly time-consuming when a hiring round attracted dozens of applications. For a batch of more than 70 CVs, the client reported that initial screening previously took approximately one week. The objective was to automate the repetitive first-pass review while preserving the ability for recruiters to inspect candidate assessments and make their own hiring decisions. I developed an n8n workflow connecting Google Drive, PDF text extraction, Gemini, and Google Sheets. Candidate documents are retrieved from Google Drive, their readable contents are extracted, and Gemini evaluates the information against criteria defined for the specific role. The resulting candidate scores, explanations, and review information are organized in Google Sheets. An important requirement was handling documents that could not be reliably extracted. Rather than generating assessments from incomplete or unavailable information, the workflow flags those files for manual review. The resulting system supports recruiters with structured initial assessments while keeping human judgment part of the recruitment process.
The workflow was designed for an existing recruitment process using Google Drive for candidate documents and Google Sheets for organizing screening information. It did not require MugDevs to adopt a separate applicant tracking platform. Gemini provided language-model-assisted assessment of readable CV content, while n8n coordinated document retrieval, processing, and result delivery. The scoring process used criteria defined for the hiring role. Its results were intended to assist initial screening, not independently determine which candidates should be hired. The workflow also included explicit handling for documents that could not be processed reliably, distinguishing unavailable evidence from a genuinely low candidate score.
Contribution
I developed an AI-powered recruitment screening workflow for MugDevs between April and May 2026. My responsibilities included n8n workflow implementation, Google Drive document intake, PDF text extraction, Gemini integration, role-specific candidate assessment, Google Sheets result organization, and handling documents requiring manual review.
My contribution focused on converting a repetitive manual screening process into a structured workflow that recruiters could use during hiring rounds. The implementation needed to do more than summarize CVs. It had to assess candidate information against defined job requirements, produce useful explanations, and organize results so recruiters could review applicants consistently. A significant part of the work involved ensuring that documents without reliable extractable content were identified rather than treated as successfully screened applications.
Designed and implemented the screening automation using n8n. Connected document intake, text extraction, AI assessment, and spreadsheet output into a coordinated workflow that could process candidate applications for a hiring role.
Integrated Google Drive as the source of candidate CVs. Built the document retrieval stage that supplies applicant files to the screening process, allowing the workflow to use the client's existing document storage.
Implemented extraction of readable text from candidate PDF documents. Prepared the extracted information for Gemini so applicants could be evaluated based on the content available in their submitted CVs.
Integrated Gemini to evaluate extracted CV information against criteria defined for the hiring role. The assessment generated candidate scores and supporting explanations, providing recruiters with structured screening information rather than requiring them to interpret raw model responses.
Connected screening results to Google Sheets. Organized candidate assessments, rankings, supporting justifications, and review status in a spreadsheet format that recruiters could use for comparing applicants and planning further evaluation.
Implemented handling for documents that could not be reliably processed, including DOC/DOCX files and image-only or scanned PDFs without usable extracted text. Directed these applications toward manual review rather than generating candidate scores without sufficient supporting information.
Delivered the automation as a paid client workflow for MugDevs. The workflow was used for recruitment screening and is currently idle between hiring rounds, with maintenance available when needed.
The project combined AI interpretation with conventional workflow automation. n8n coordinated the process, Google Drive supplied candidate documents, Gemini evaluated readable content against role-specific criteria, and Google Sheets provided the review interface. The implementation deliberately distinguished between applications that could be assessed using available CV text and those requiring manual inspection. The resulting automation reduced repetitive initial screening work without replacing recruiters' responsibility for reviewing candidates or making hiring decisions.
Implementation
The AI CV Screener was implemented as an n8n automation connecting Google Drive, PDF text extraction, Gemini, and Google Sheets. The workflow processes candidate documents, evaluates readable CV content against criteria defined for a particular role, and generates structured screening results. The implementation separates document retrieval, text extraction, AI assessment, and result organization so each stage performs a specific responsibility. It also accounts for unsupported or unreadable documents by directing them toward manual review instead of treating incomplete information as a valid candidate assessment.
Integrated Google Drive into the n8n workflow to retrieve candidate CVs submitted for screening.
The workflow uses the stored documents as the source of applicant information, allowing recruiters to continue working with their existing file-storage system.
Retrieved files are passed to the document-processing stage for extraction and subsequent assessment.
Implemented a document-processing stage to extract readable information from candidate PDF files.
The extracted text provides the information used by Gemini when assessing candidates against role-specific requirements.
The implementation distinguishes documents containing usable text from files that cannot be reliably interpreted through the available extraction process.
This is important because a valid candidate assessment requires sufficient information from the submitted document.
Integrated Gemini to evaluate extracted CV content against criteria defined for the hiring role.
The model analyzes the available candidate information and generates structured assessments that include scores and supporting explanations.
Using role-specific criteria makes the evaluation relevant to the position being screened rather than relying on a single generic set of qualifications for every application.
The resulting assessments serve as initial screening information for recruiters.
Organized the model-generated assessments into a consistent structure that could be used by the rest of the automation.
The workflow processes screening results so candidate scores and supporting explanations can be recorded in defined spreadsheet fields.
This makes the assessments easier to review and compare than unstructured text responses.
The structure also provides a distinction between completed assessments and applications that require additional human inspection.
Integrated Google Sheets as the destination for candidate screening information.
The workflow writes assessment results, including scores, justifications, rankings, and review status, into a spreadsheet that recruiters can use during the hiring process.
This provides a centralized view of the screened applications without requiring the client to work directly with n8n execution data or raw Gemini responses.
The spreadsheet serves as a review aid rather than an automated hiring decision system.
Not every submitted CV could be processed reliably using the available extraction workflow.
DOC/DOCX documents and image-only or scanned PDFs without usable text were handled as cases requiring manual review.
Rather than generating AI scores from missing or incomplete evidence, the workflow identifies these documents for human inspection.
This avoids presenting unsupported assessments as if they were based on successfully extracted candidate information.
Delivered the connected automation for MugDevs as a paid client engagement.
The system brings together candidate document intake, readable text extraction, Gemini-based assessment, and spreadsheet output.
Following delivery, the workflow entered an idle state between recruitment rounds rather than continuously processing applications.
Maintenance is provided as needed.
Outcomes
The AI CV Screener significantly reduced the manual effort required for MugDevs' initial recruitment screening. For a hiring batch involving more than 70 CVs, the client reported that screening time decreased from approximately one week to minutes. The workflow also introduced a structured approach to assessing candidates against defined role requirements, with scores and supporting explanations recorded in Google Sheets. Applications that could not be reliably extracted were directed toward manual review rather than receiving unsupported AI-generated scores. These outcomes are based on the delivered client workflow and reported screening experience. No claims are made about hiring accuracy, elimination of recruitment bias, or independently measured model performance.
Outcome
Automated the initial review of candidate CVs, reducing the client's reported screening time for a batch of more than 70 applications from approximately one week to minutes.
Measured impact
Previously approximately one week for more than 70 CVs.
The improvement was reported by MugDevs following use of the workflow. It describes initial screening time, not the complete recruitment or hiring process.
Deliverable
Implemented Gemini-powered screening that evaluates readable CV content against criteria defined for the hiring position.
The workflow generates structured assessment scores and supporting explanations to help recruiters compare applicants using the same defined criteria.
Deliverable
Connected Google Drive document retrieval with PDF text extraction, reducing the need for recruiters to open and manually review every readable CV during the initial screening stage.
Candidate documents are processed through the automation before their extracted information is supplied to Gemini for assessment.
Deliverable
Integrated Google Sheets to organize candidate scores, rankings, assessment justifications, and review information in a familiar spreadsheet format.
Recruiters can review the resulting information without needing to inspect individual workflow executions or raw AI responses.
Deliverable
Implemented handling for unsupported document formats and image-only PDFs, directing those applications toward manual review rather than producing unreliable candidate scores.
This preserves a distinction between candidates assessed using readable evidence and applications that require additional inspection.
Outcome
Established a repeatable screening workflow using role-defined evaluation criteria instead of relying entirely on varying manual first-pass reviewer judgments.
Consistent criteria provide a common basis for organizing initial assessments. This does not establish that AI-generated scores are unbiased, accurate, or suitable for making final hiring decisions without human review.
Outcome
Completed and delivered the recruitment screening automation as a paid client engagement for MugDevs.
The engagement ran from April to May 2026. The workflow is currently idle between hiring rounds, with maintenance provided as needed.