EXPERIMENT OVERVIEW

The VAFER experiment will deploy an innovative onboarding and training support system for high-tech manufacturing, leveraging Artificial Intelligence (AI) and the Open Voice OS (OVOS) framework. The experiment involves R2M Solution, the Digital Innovation Hub Lombardia (DIHL), and the T&G electronic repair centre for real-world validation.

The solution operates in two main phases to improve human-machine collaboration. First, it conducts a precise skill assessment by combining automatic CV analysis via R2M’s AVANTI software with a natural, voice-driven AI interview powered by OVOS to identify exact trainee competencies and eliminate redundant training. Second, it provides a Digital Intelligent Assistant (DIA) for real-time, interactive voice assistance during practical training sessions. This will be demonstrated by allowing trainees to receive step-by-step guidance, instructions, checklists and risk explanations through voice interaction while keeping their hands and eyes focused on electronic boards. This experiment is highly relevant to the WASABI ecosystem because it combines Large Language Models (LLMs) and OVOS voice framework in AI agents that considerabily improve operational efficiency in real cases.

The primary challenge that VAFER addressed is the expensive and time-consuming nature of the onboarding process, especially in high-tech manufacturing and electronic repair. Currently, new hires need to read extensive operational procedures, study complex equipment manuals, and spend considerable time training on a physical workbench. This scenario has many pain points: companies view onboarding as a financial burden and productivity drain (although necessary), while trainees may experience high mental load and frustration from redundant skill assessments and the pressure of memorizing lengthy checklists. The second use case, more operational and technical, is highly sensitive in a context like T&G. Integrating new employees disrupts established workflows, and errors made during delicate operations—such as repairing Printed Circuit Boards (PCBs)—can severely damage equipment or compromise safety protocols. A standard chatbot could help, but writing questions with a keyboard obliges to look away, reduces attention, and can ultimately lead to mistakes.

Objective 1: 

Reduce average onboarding time by up to 15% compared to the current baseline.

  • This will be achieved by utilizing the AVANTI solution and the OVOS-powered Digital Intelligent Assistant to assess existing skills and allow mentors to remove redundant training modules

Objective 2:

decrease average rate of errors during practical training sessions by up to 25%.

  • The OVOS AI agent will provide on-demand, hands-free guidance, for example reminding correct sequence of actions and explaining potential risks. Operators will then receive support while keeping visual and manual focus on workbench tasks. To further reduce risks of damages to PCBs, we will also provide a VR-powered, virtual workbench where new hires will exercise in a no-stress, zero-risk environment.

Objective 3: 

improve trainee onboarding satisfaction by up to 20% and enhance training retention by up to 15%. 

  • The preliminary interaction will lead to a reduction or removal of topics where trainees are already proficient, reducing frustration.

 

We will carry out the experiments in the electronic manufacturing and repair sector, specifically focusing on the onboarding and on-the-job support of personnel. To validate the use case, we will rely on our pilot T&G Repair and Remanufacturing Centre in Italy, a facility that specializes in extending the lifecycle of obsolete electronic components and Printed Circuit Boards (PCBs), processing around 13,000 repairs annually. Target users include new hires undergoing training (first scenario), existing employees requiring reskilling (second scenario), and HR professionals or mentors overseeing the process (both).

Key constraints include managing the varying quality of input data (such as unstructured CVs and dense company manuals), ensuring stringent data privacy and security, and safely integrating the system within an environment that features complex procedures, specialized machinery, and strict safety protocols.

EXPECTED IMPACT

EXPECTED IMPACT

The success of the VAFER experiment will be measured through concrete Key Performance Indicators (KPIs) tracked against the company’s current baselines. The measurable targets include a 15% reduction in onboarding time, a 20% improvement in user onboarding satisfaction (measured via pre- and post-assessment surveys), a 25% reduction in errors during practical sessions, and a 15% increase in training retention (measured via mentor assessments).

End users will benefit from a personalized, engaging training experience that drastically reduces mental burden, while stakeholders will see reduced training costs and smoother workflow integration. Environmentally, by successfully training operators to repair and remanufacture obsolete electronic components rather than discarding them, the solution directly supports sustainability and circular economy principles in the electronics lifecycle.