Customer

A leading technology company in consumer computing, known for pushing the boundaries of user experience through AI-driven innovation.

Problem Statement

The customer aimed to evolve its existing AI chat-based troubleshooting system. While the earlier version could answer user’s queries using manuals and system documentation, it could not take corrective actions.

The next step was to explore enabling the AI agent to directly execute troubleshooting tasks using the customer’s pre-installed diagnostic and utility tools on laptops. This prototype capability was intended to reduce user effort, minimize downtime, and move toward a more seamless support experience.

Solution Overview

MulticoreWare enhanced the customer’s AI troubleshooting system by integrating early-stage action-execution capabilities into the existing conversational framework, gradually transforming it from a passive support tool into a more active, self-healing–oriented prototype agent.

Key aspects of the solution include:

1. Function Calling & Tool Invocation: Extended the RAG-powered agent with secure function calling to invoke the customer’s built-in utilities such as diagnostics, battery health checks, driver updates, and network resets.
2. SLM Fine-Tuning: Fine-tuned Gemma 2B to more accurately map natural language troubleshooting queries into system-level actions using the customer’s diagnostic and utility APIs.
3. On-Device Execution: Deployed the advanced prototype on AI PCs with integrated NPUs, enabling low-latency, offline-capable operation.

This allows the AI agent to autonomously perform corrective actions or guide users interactively through the troubleshooting process, streamlining issue resolution and elevating user satisfaction.