AI-Powered CV Scoring System
Overview of Our Client
Our client is a company with a high volume of ongoing recruitment activities, regularly processing large numbers of applications across multiple roles. As their business grew, so did the number of incoming CVs, creating increasing pressure on the HR team.
Manual review required significant time and often resulted in inconsistent evaluations and a higher risk of errors.
To keep up with hiring demands while maintaining fairness and compliance, the client was looking for a way to streamline and standardize the early-stage candidate review process without compromising data privacy or regulatory requirements.
Challenge
Despite the growing adoption of AI in recruitment, the client faced several operational challenges that reduced the efficiency of their hiring workflows. Recruiters had to manually review an ever-increasing number of CVs, which made the initial screening stage time-consuming and slowed down the overall recruitment cycle.
At the same time, subjective human assessments often resulted in inconsistent candidate scoring, while the heavy workload increased the risk of overlooking strong applicants. In addition, processing personal data introduced strict compliance and privacy requirements that further complicated the process.
Main Goals
- Automate the initial screening and scoring of large volumes of CVs.
- Ensure consistent and criteria-based evaluation across different roles.
- Enable batch processing for faster candidate review cycles.
- Introduce guardrails to automatically remove personal and sensitive data.
- Maintain high scoring accuracy through testing and evaluation workflows.
- Provide an architecture that can be easily integrated into existing HR systems.
Project Overview
We built an AI-powered system that helps companies evaluate large volumes of CVs more efficiently. The platform automatically processes resumes, extracts relevant information, and evaluates candidates based on predefined criteria for each role.
The platform supports batch CV processing, allowing recruiters to review many applications faster. It also includes automated mechanisms that detect and remove sensitive personal data to support privacy and compliance requirements.
- Industry: Recruitment / HR Tech
- Region: Europe
- Project Duration: 1 month
Solution
We delivered an AI-powered CV scoring software designed to automate candidate evaluation during HR processes. The system uses LLMs to analyze resumes against predefined criteria for each position and generate consistent scores.
The solution processes CVs in batches and includes built-in guardrails that remove sensitive and personal data before results are provided. It was implemented using LangChain for orchestration, OpenAI models for evaluation, Docling for document processing, and MLflow to support testing and accuracy monitoring.
Key Features
- Automated CV scoring based on role-specific criteria
- Batch processing of large volumes of resumes
- LLM-driven evaluation workflow
- Guardrails for removing personal and sensitive data
- Testing workflows to maintain high-scoring accuracy
Technology Stack
To implement the CV scoring solution, we used the following technologies and tools:
LLM Orchestration
- LangChain
LLM Provider
- OpenAI
Document Processing
- Docling
Model Tracking & Evaluation
MLflow
Related Cases
Core Team
- Solution Architects: Designed the overall CV scoring architecture and evaluation workflow.
- AI Engineers: Implemented LLM-based scoring logic, criteria evaluation, and guardrails for data privacy.
- Backend Developers: Built CV processing pipelines and system integrations.
- QA Engineers: Tested scoring accuracy and validated system performance.
Results
The delivered solution introduced an AI-powered CV batch processing workflow that significantly reduced the need for manual resume screening and enabled automated AI resume screening at scale. Candidates are now evaluated automatically against role-specific criteria, enabling faster and more consistent hiring decisions.
The platform also ensures privacy compliance through built-in guardrails that remove personal data, while testing workflows help maintain high-scoring accuracy over time.