Building a Robust AI-Powered E-learning Tool for a German Automaker

Many well-established businesses in the traditional industries, including the automotive sector, struggle with the migration of their old, legacy infrastructures to the cloud, as well as their optimization for robust service provision in line with present-day business needs. Here is a case study of the RNDpoint team’s collaboration with a large European automaker on the enhancement of its internal e-learning and HR screening tools with the help of AI technology. 

Client’s Profile 

The Client that turned to RNDpoint for assistance is a large-scale German automaker. 

Client’s Requirements and Goals 

The Client approached RNDpoint with the requirement to design an AI-powered recommendation system that would enhance the company’s recruitment and onboarding, and expand the internal e-learning system’s capabilities. The requirements and goals were as follows: 

• A sophisticated recommendation system for the Client’s internal e-learning platform powered by an advanced LLM. 
• An AI-powered interviewer system capable of analyzing candidate CVs in PDF format and selecting the top-matching contenders in compliance with a multi-parameter screening checklist. 
• An internal e-learning AI assistant able to simplify onboarding process, respond to user queries and generate content in response to user prompts. 
• LLM training with platform-specific data coupled with model quantization techniques for the model’s optimal efficiency. 
• The e-learning and onboarding software migration to cloud services for infrastructure scaling and compliance with increasingly complex requirements. 

Business Challenge 

The RNDpoint team encountered several business challenges in the process of working on this LMS for manufacturing. They included: 

• Data processing. The designed solution had to effectively process heterogeneous candidate CV data and extract relevant information with a high degree of accuracy to make correct competency and fit evaluations.  
• Candidate evaluation. RNDpoint engineers had to develop sophisticated algorithms for candidate ranking across multiple competency parameters, including relevant skills, experience, and qualifications.  
• Content and response generation. Our task was to create an AI assistant that could effectively navigate platform-specific data and give users accurate, relevant, and quick responses to their queries on the internal e-learning platform. 
• LLM training. The training process should be conducted on the automaker’s proprietary platform data, with a high level of efficiency guaranteed by model quantization. 
• Infrastructural scalability. RNDpoint was tasked with migrating the Client’s existing e-learning and HRM screening solutions from local GPU servers to the cloud to ensure a high level of infrastructure scalability. 

Implementation 

The RNDpoint team worked on the following aspects of the project during its partnership with the client. 

AI Interviewer System Design 

A robust AI-powered interviewer system was developed to enable efficient data extraction from candidate CVs in PDF format using advanced parsing technologies. AI candidate ranking algorithms were put in place, with custom algorithms for evaluation and ranking against numerous parameters (e.g., technical skills, work experience, educational credentials, and qualifications relevant to each position). The system was equipped with OpenAI models and HuggingFace embeddings to facilitate its understanding of natural language. LLM integration was performed using Llama-2 and OpenAI models for extracted data analysis. A sophisticated ML model was developed to predict the candidate’s suitability for specific roles with a high degree of accuracy. 

AI Assistant on the E-Learning Platform 

The AI assistant was created on the basis of Langchain and LlamaIndex, which enabled its conversational interactions with users and educational content generation in response to user queries. The algorithm was further trained on platform-specific data to give accurate, contextually appropriate responses in the form of data summaries, explanations, and hand-picked learning materials. OpenAI Embeddings enabled proper semantic understanding and accurate content retrieval. Continuous learning mechanisms were implemented to enable the AI model’s updates with new information loaded to the platform and content generation using it. 

LLM Training and Optimization 

The LLMs were trained with the Client’s proprietary data, with curated datasets derived from the platform for comprehensive model training. We applied model quantization techniques to reduce the model’s complexity and enhance its processing capacity, which resulted in increased inference speed. As a result, we’ve balanced efficiency improvements with high accuracy, performing the entire process in line with data privacy and compliance standards. 

Infrastructure Management, Scaling, and Cloud Migration 

AI solutions developed on the Client’s order were initially deployed on the local GPU servers, with local resources utilized to manage computationally demanding tasks in the development process. This approach enabled RNDpoint developers to complete rapid prototyping and testing. After that, the whole system was transferred to Azure Cloud Services for the sake of building scalable cloud infrastructure instead of the Client’s legacy system. RNDpoint leveraged Azure’s machine learning and data processing services to improve the platform’s reliability and high throughput. 

Solution Delivered 

The solution produced by RNDpoint represents a comprehensive AI-powered, cloud-based system for internal e-learning and automated CV screening with the following features: 

• An AI-powered algorithm that processes CVs and ranks candidates in line with the set parameters. This innovation has enabled more effective application processing and an enhanced recruitment process due to data-backed insights into candidate qualifications. 
• An advanced e-learning assistant that improves user experience and provides instant guidance and responses to user queries with GenAI capacity. 
• A customized LLM trained on the Client’s internal data and exhibiting improved model performance due to quantization and optimization techniques. 

Besides, the RNDpoint team has completed the Client’s e-learning system migration to Azure cloud to ensure seamless scaling of its infrastructure in line with dynamic demand. The migration process was accompanied by frictionless integration of Azure with the existing systems with minimal downtime in the system’s operations. 

Project Outcomes 

The LMS for employee training within manufacturing company took six months to complete, from January 2023 to June 2023. Five RNDpoint specialists worked on the task, including three AI engineers tasked with developing and fine-tuning the required language models, one QA engineer, and one Project Manager who coordinated communication between the Client’s stakeholders and the development team. 

The technology stack used in this project covered OpenAI language models and embeddings, Llama-2, HuggingFace Embeddings, Llama Index, Langchain, and Azure Cloud Services. The resulting business solution represents a scalable and robust e-learning and HRM infrastructure that successfully automates target activities within the company and seamlessly integrates with existing systems.