New Empire
Case Study
← All work

An AI-powered SaaS that automates the busywork

Built a Generative-AI SaaS product that automated 60%+ of manual workflows using OpenAI, RAG, and a cloud-native backend.

RoleSenior Software Engineer & Technical Lead
Year2024
StackOpenAI · RAG · Node.js · Python · AWS · PostgreSQL
60%+Workflows automated
Multi-LLMOpenAI · Gemini · HF
RAGGrounded answers

The challenge

  • Teams were drowning in repetitive, manual knowledge work that didn't scale with headcount.
  • Off-the-shelf AI tools hallucinated and couldn't be trusted with real business context.

The approach

  • Designed a Retrieval-Augmented Generation pipeline so answers were grounded in the customer's own data.
  • Integrated multiple model providers (OpenAI, Gemini, Hugging Face, vLLM) behind a single abstraction for cost and quality control.
  • Wrapped it in a production SaaS with auth, billing-ready architecture, and observability.

The outcome

  • Automated more than 60% of previously manual workflows.
  • Delivered trustworthy, source-grounded AI output instead of generic chatbot guesses.
  • Created a repeatable AI product foundation the business could extend to new use cases.