Automated CV Screening and Evaluation with Local LLMs and Google Sheets - n8n Workflow

HR Managers and Recruitment teams needing rapid, structured CV shortlisting.
Technical users seeking powerful n8n templates for self-hosted AI automation using Ollama.
Businesses that require standardized, quantifiable candidate evaluation based on dynamic criteria.
Developers looking for an advanced example of using the n8n node ecosystem for document processing and AI.

This sophisticated n8n workflow solves the challenge of manual, time-consuming candidate screening by automating the entire evaluation process. It integrates cloud storage (Google Drive) and data management (Google Sheets) with a highly customizable local AI model (Llama 3.1, run via Ollama). The n8n automation extracts candidate data, applies dynamic evaluation criteria fetched from a specific Google Sheet, and uses the LLM to generate precise, structured scores and ratings. The final output is compiled back into a Google Sheet, offering a clear, quantifiable comparison of all candidates. This robust n8n workflow dramatically increases efficiency and standardization in high-volume hiring scenarios, utilizing powerful AI capabilities delivered through a repeatable n8n node sequence.

The n8n workflow begins execution when the Webhook n8n trigger receives an external signal, initiating the batch CV processing job.

1. Search and Loop: The Folder-search n8n node locates all target CV files within a designated Google Drive folder. The Loop Over Items n8n node then processes these files iteratively, handling one candidate at a time.


2. Data Ingestion: Inside the loop, the Download-file node pulls the PDF CV, and the Extract from PDF n8n node converts its binary content into readable text. Simultaneously, the Input-criteria n8n node fetches the current evaluation matrix (matching criteria and scoring output type) from Google Sheets.


3. Context Assembly: The CV text and the criteria are merged using the Merge and Aggregate n8n nodes, preparing the full context for the AI.


4. AI Analysis: The Basic LLM Chain constructs a detailed prompt using the extracted CV text and the criteria. It sends this to the Ollama Model running Llama 3.1, leveraging the Structured Output Parser to enforce the output format. This ensures the AI provides a reliable, structured JSON evaluation (including candidate name, file name, scores for each criteria, and overall points).


5. Data Transformation: The Code n8n node performs crucial post-processing. It dynamically gathers all unique evaluation criteria fields generated by the LLM and reformats the structured JSON output into flat, standardized rows suitable for spreadsheet insertion, making this n8n workflow flexible to changing criteria.


6. Final Output: The Output-sheet n8n node appends the comprehensive, scored evaluation results to the final tracking Google Sheets document, completing the automated screening process for that candidate. This makes the entire n8n workflow reusable and efficient.

1. Import: Import this n8n workflow JSON into your n8n instance.


2. Ollama Credentials: Set up credentials for the Ollama Model n8n node, pointing it to your local or remote Ollama instance where Llama 3.1 is installed.


3. Google Drive/Sheets Credentials: Configure OAuth2 credentials for Google Drive and Google Sheets. Ensure these accounts have read access to the input criteria sheet and CV folder, and write access to the output sheet.


4. Configuration: Update the Folder-search node to specify the ID of your Google Drive folder containing the CVs.


5. Google Sheets Setup: Update the Input-criteria and Output-sheet n8n nodes with the correct Document IDs and Sheet Names for your criteria template and results log, respectively.


6. Activate: Activate the n8n workflow and trigger the initial Webhook n8n trigger to start processing.

Webhook (n8n trigger): The starting point of this automation, providing an external API endpoint to initiate the n8n workflow run.
Google Drive (Folder-search, Download-file): Handles file operations. Searches a specific folder ID for files, then downloads individual files based on the iteration from the Loop Over Items n8n node.
Google Sheets (Input-criteria, Output-sheet): Manages external data. Input-criteria pulls the template of questions used for AI evaluation. Output-sheet appends the final results.
Extract from PDF: Converts the downloaded PDF CV files into raw text, which is essential input for the LLM chain.
Ollama Model (llama3.1:latest): The core AI engine. This n8n node sends the comprehensive prompt to the local Llama 3.1 model, ensuring the AI performs the complex scoring analysis.
Basic LLM Chain: Constructs the dynamic prompt by injecting candidate text and evaluation criteria, instructing the LLM on the required analysis and desired output structure. It relies on the downstream parser n8n nodes.
Structured Output Parser / Auto-fixing Output Parser: A critical pair of n8n nodes that enforce the LLM's output to strictly conform to a defined JSON schema, improving data reliability before transformation.
Code: A custom JavaScript n8n node that dynamically maps the AI-generated structured data into a flat array structure, ready for insertion into Google Sheets, ensuring the final n8n template is adaptable to various criteria.