A system that aligns external sources and proprietary data to generate player profiles with images, rankings, history and forecasts.
The problem
Producing one profile manually took about 1520 minutes, making the catalogue difficult to scale.
The solution
The pipeline separates data retrieval, entity alignment and normalization from text generation so the context remains verifiable.
What I built
I built a workflow that aligns player data, prepares structured context and generates a publication-ready editorial profile.
How it works
Player data ? entity alignment ? enrichment ? structured prompt ? LLM generation ? editorial output.
Engineering choices
The model works from structured input; data retrieval and application checks remain separate from generation.
Outcome
Average production time dropped from 1520 minutes to roughly 1530 seconds.
Gallery
