Gen AI Chatbot and Pipeline Using Vertex AI and GCP for World’s Leading Semiconductor Manufacturer
Client
The client is a global infrastructure technology leader that designs, manufactures, and supplies a wide range of semiconductor and software products. Their offerings serve multiple markets including data centers, networking, enterprise software, broadband, wireless, storage, and industrial applications.
Project Context
The client's vulnerability analysts manually researched vulnerabilities from multiple tools and websites, which was time-consuming and repetitive. They needed an automated assistant to streamline remediation recommendations and tracking.
Challenges
Analysts had to cross-check data from various tools (BlackDuck, Lacework, JFrog, Coverity, Qualys) and external web sources. The manual process delayed response times and consumed analyst effort.
Solution
SquareShift developed a chatbot powered by Vertex AI’s LLMs like Gemini Pro and Palm 2. It used a Retrieval-Augmented Generation (RAG) architecture over GCP BigQuery. A machine learning pipeline was also implemented to bulk-generate remediation suggestions and status updates.
Project Objectives
Develop a Gen AI chatbot to support vulnerability analysis using Vertex AI LLMs.
Build a pipeline that automates remediation suggestions and status handling.
Centralize data sources like Armorcode and external advisories into one actionable system.
Solution Delivery
Data Engineers scraped vulnerability-related information from various external sources and ingested it into BigQuery along with Armorcode data.
SquareShift’s AI Developer then built a Gen AI chatbot interface that interacted with this unified dataset. The chatbot not only assisted analysts with contextual responses but also interfaced with a backend ML pipeline that pushed remediation actions and status updates in bulk.
The setup significantly boosted the daily productivity of vulnerability analysts.
Testimonial
SquareShift’s Gen AI chatbot has revolutionized how we handle vulnerability management with faster responses, better accuracy, and real automation
