This repository contains a production-grade AI support agent for SpotifyCares, developed as a take-home assignment for a software development engineer intern position. It includes an offline, self-contained environment for evaluating the agent's performance against baseline models.
From the README
Production AI Support Agent for @SpotifyCares Hiver SDE Intern Take-Home Assignment: An industrial-grade, grounded customer support AI agent for @SpotifyCares built on real-world Twitter customer support conversations. --- ? Quickstart: Reproduce Headline Results in < 1 Minute This repository is 100% self-contained and runnable offline with zero mandatory paid API keys or external credentials required. 1. Installation 2. Run Headline Evaluation Benchmark (< 5 seconds) This executes the automated evaluation harness comparing the Production Agent against the Simple Baseline and Trivial Baseline across all 200 Golden Evaluation examples and 50 Human Judge benchmarks. 3. Run Automated...
Skills
- Python
