AI HR Assistant for Secure and Efficient Enterprise Recruitment

HR Tech

Isolated AI

AI Automation

ai hr assistant for secure and efficient enterprise recruitment

Project overview

Ready to see what we can do for you?

Industry

HR Tech / Recruitment Automation

Client Location

Europe (undisclosed enterprise client)

Recognition

Internal pilot of Evinent Private AI for enterprises

Provided Services

Development of isolated AI agents for HR workflows

Type of the Project

Proof of Concept (PoC) / Pilot implementation

Collaboration Model

Fixed-scope pilot with agile iterations

Duration

4–6 weeks (setup, configuration, testing)

2 AI engineers, 1 PM/BA

Team Size

Project business objectives

Automate candidate-vacancy matching within a large hiring system.

Create separate AI agents for recruiters and candidates.

Ensure complete data isolation — no external API calls to OpenAI, Claude, or Gemini.

Enable fast deployment and minimal infrastructure load.

Maintain customizable logic and s for each department or user role.

Test feasibility of Evinent’s Private AI for Enterprises concept in HR use cases.

About the project

A private AI solution developed by Evinent to automate HR workflows in a large retail company. The project focused on creating isolated AI agents that assist both recruiters and candidates in finding the right match across thousands of vacancies, while ensuring complete data privacy and flexibility in deployment.

Process overview

We used an iterative development approach, where we slowly progressed through building and implementing different parts of the chatbot. Through this platform, the client could test the functionality early on, make the necessary adjustments without significant rework, and adjust the system to their business needs.

Discovery & Requirements Alignment

Data structure review

Agent Setup

Prоmpt & Business Logic Customization

Testing & iteration

Pilot deployment

Discovery & Requirements Alignment

Analysis of recruitment workflows and user roles (recruiter vs candidate).

Data structure review

Connection to existing vacancy and candidate databases.

Agent Setup

Creation of two AI agents:

Recruiter Assistant — searches candidates, filters by experience, skills, language, and availability.

Candidate Assistant — helps applicants find the most relevant open positions based on experience and preferences.

Prоmpt & Business Logic Customization

Defining response templates and fallback behavior when no results are found.

Testing & iteration

Validation of matching logic, handling of edge cases, internal QA.

Pilot deployment

Controlled rollout inside HR department sandbox.

Discovery & Requirements Alignment

Analysis of recruitment workflows and user roles (recruiter vs candidate).

Data structure review

Connection to existing vacancy and candidate databases.

Agent Setup

Creation of two AI agents:

Recruiter Assistant — searches candidates, filters by experience, skills, language, and availability.

Candidate Assistant — helps applicants find the most relevant open positions based on experience and preferences.

Prоmpt & Business Logic Customization

Defining response templates and fallback behavior when no results are found.

Testing & iteration

Validation of matching logic, handling of edge cases, internal QA.

Pilot deployment

Controlled rollout inside HR department sandbox.

Challenges we faced and how we overcame them

Ready to see what we can do for you?

Сhallenge 1

Risk of hallucinations and inconsistent responses in chat-based interaction.

solution

To ensure reliability and consistency, we implemented an atomic agent architecture, where each agent was responsible for a specific, well-defined task such as searching, matching, or summarizing data. This eliminated overlapping logic and reduced the risk of misinterpretation or “hallucinations.” By isolating responsibilities, the system maintained predictable and verifiable outputs, making the AI’s behavior easier to monitor and fine-tune.

Сhallenge 2

Integration with existing HR databases and variable data quality.

solution

The client’s data sources included legacy formats and inconsistent entries, which complicated automated search and matching. Our team developed custom API connectors and data normalization scripts to clean, standardize, and structure the HR data before feeding it into the AI search index. This approach improved the accuracy of candidate-vacancy matching and ensured that even incomplete records could be processed effectively without manual intervention.

Сhallenge 3

Security and compliance expectations.

solution

As data privacy was a top priority, we deployed each AI agent within isolated containers with role-based access control to separate user environments and permissions. No external API calls or third-party data transfers were used; all processing occurred within the client’s infrastructure. This setup minimized exposure risks and aligned with internal compliance requirements for sensitive recruitment data.

Key security measures

Ensuring the privacy of recruitment data and protecting internal information from potential leaks was a top priority for the client.Evinent implemented a robust security framework specifically designed for isolated enterprise AI environments, ensuring full compliance with corporate data protection policies and internal IT governance.

Isolated Deployment Environment

Each AI agent operates in its own containerized environment, fully separated from external systems. This guarantees that no data leaves the client’s infrastructure and eliminates dependency on third-party LLM APIs.

Role-Based Access Control (RBAC)

Access to the AI system is managed through defined user roles and permissions, allowing HR managers, recruiters, and administrators to work securely within their assigned scopes.

Data Encryption and Storage Policies

All communication between modules and data endpoints is encrypted. Sensitive recruitment data, such as candidate resumes and vacancy details, is processed only within secure internal networks.

Custom Logging and Monitoring

The system can be extended with logging modules to track agent activity, query types, and system performance. This ensures full transparency and supports security audits if required by the client.

Compliance-Ready Architecture

While no formal certifications were required at the pilot stage, the solution follows security best practices compatible with GDPR and ISO 27001 frameworks. This ensures the architecture can easily pass compliance audits when scaled to production.

Data protection compliance

Evinent’s isolated AI framework ensures data never leaves the client’s secure perimeter. By combining containerized architecture, access control, and internal encryption policies, the HR AI assistant fully complies with enterprise-grade data protection and privacy standards while maintaining the flexibility to scale across other departments.

Technology stack

AI & NLP

Models: Multiple open-source LLMs with permissive commercial licenses (Apache, MIT)
Capabilities: Text classification, semantic search, summarization, and context-based matching
Approach: Atomic agent architecture, where each agent handles a focused task (search, match, summarize) for maximum precision and control

Frontend

Languages & Frameworks: JavaScript, TypeScript
Frameworks: Angular
Interface Purpose: Internal HR dashboard for managing agents, monitoring activity, and adjusting parameters

Backend

Languages & Frameworks: C#, .NET, Web API
Integration Layer: Custom REST APIs for communication between AI agents and HR databases
Logic Management: Configurable business rules and templates stored per department or user role

Database & Data Storage

Databases: PostgreSQL
Search & Indexing: Elasticsearch for fast and relevant candidate–vacancy matching
Data Processing: Custom connectors and normalization scripts to clean and standardize legacy datasets

Infrastructure

Containerization: Docker-based isolated containers for each AI agent
Deployment Options: On-premises, private cloud, or hybrid configurations based on client security policy
Access Management: Role-Based Access Control (RBAC) and internal authentication

Security & Compliance

Data Protection: Encrypted data flow between modules; no external API calls or data transfers
Isolation: Each AI instance runs in a secure, independent environment
Compliance Readiness: Architecture aligned with GDPR and ISO 27001 requirements

Monitoring & Performance

Monitoring Tools: Optional integration with Grafana or internal dashboards for usage analytics
Logging: Configurable activity logs and system health monitoring for audit and performance tracking

Features

Multi-model selection

different LLMs used for search, summarization, and communication tasks.
Fallback logic for “no results found” scenarios (clarify criteria / propose alternatives).

Role-based access

each HR specialist or manager has isolated AI instances.
Configurable parameters: temperature, tone, stop words, and search precision.

Customer satisfaction metrics

HR team reported faster candidate filtering and higher relevance in shortlists.

Reduced manual workload in vacancy-candidate pairing.

Consistent AI behavior without hallucinations thanks to the atomic architecture.

Impact on company’s business growth

Evinent’s isolated AI solution helped the client improve HR efficiency and data security while laying a foundation for scalable automation across departments:

Reduced Screening Time

The AI assistant significantly accelerated candidate filtering and matching, allowing HR teams to process large candidate pools faster.

Improved Data Utilization

Clean, structured, and searchable HR data enabled more accurate vacancy-candidate matches and better internal reporting.

Lower Operational Costs

Automation of routine tasks minimized manual workload and reduced dependency on external AI tools or paid API tokens.

Scalable Architecture

The modular, containerized setup created a replicable model for future AI deployments across other corporate functions such as customer support and internal documentation management.

Project results

Working prototype of recruiter and candidate AI assistants.

Configured environment with isolated containers for each agent.

Custom API connectors for HR databases.

Documentation and recommendations for scaling into production.

Client feedback

For us, privacy and stability were key. Evinent’s team delivered a working AI assistant that genuinely understands our recruitment data without sending it anywhere outside. The setup was surprisingly fast, and we’re already discussing expansion to other internal workflows.

Next project

Explore all case studies
Explore all case studies

Secure data synchronization and legacy migration for a healthcare client

Explore how Evinent implemented a scalable and secure data infrastructure for a leading US healthcare technology provider. The solution included advanced encryption, real-time data synchronization, and modernization of legacy systems, ensuring efficient operations and compliance with stringent security standards.