2

GenAI - ask the docs

Create an ‘ask-the-docs’ GenAI tool that could be hosted on GCP for multiple users to access.

Description: There was an ask and need for PHAC to start looking into building an ‘ask-the-docs’ type of application that could be stored on Google Cloud Platform (GCP) in a secure way which would allow users to upload or read from local documents and query them with a chat-like interface/feel. Multiple tools could do such a thing and it was my responsibility to learn them, deploy them, and test them to determine which would meet the needs of the team(s).

Roles and Responsibilities: Full-stack and Cloud Developer

  • Researched which current methods exist and narrowed down options
  • Deployed the options within a GCP environment (using GenAI tools such as VertexAI and DocAI)
  • Deployed some local LLMs (using PrivateGPT and Mistral, LLama, OpenAI, and HuggingFace)
  • Tested each deployment’s capabilities and use cases along with performance and UAT testing
  • Created documents and presentations to brief senior management on the findings and provide recommendations based on the results of the tests

Project Duration: 1 year

Environment:

  • Git Bash
  • Python
  • git
  • GCP Cloud SQL
  • GCP Container Registry
  • GCP Kubernetes clusters
  • GCP IAM
  • GCP GenAI
  • GCP DocAI
  • PostgreSQL
  • LLMs
  • PrivateGPT
  • OpenAI
  • .docx
  • .xlsx
  • .csv