Tridge · April 2025–present
AI-Powered News Classifier
Assistant Manager, ML Engineer · Backend Department AI Team
LLM news classification and backend processing at Tridge.
The problem
Categorize news content using language models while improving processing speed and backend system efficiency.
My contribution
- Develop the LLM-based news classifier.
- Test multiple model versions and improve processing algorithms.
- Contribute to the performance and efficiency of company systems.
Technical approach
The work connects model comparison with backend processing. Classification quality and processing speed are considered together when evaluating model versions and algorithm changes.
Model comparison
Test multiple language-model versions for news classification.
Processing
Improve the algorithms that process content within the backend system.
Evaluation focus
Consider classification quality alongside processing speed. Public benchmark records are not available for this work.
Engineering considerations
Faster processing is only useful if classification remains useful. Model comparisons need consistent labels and evaluation data; aggregate accuracy alone can conceal weak performance on less common categories.
Outcome
An ongoing news-classification implementation and model-optimization effort within the backend team.
Evidence & scope
This is a professional-work summary. Internal code, evaluation datasets, and benchmark results are not included in this public portfolio.
Project resources
This page summarizes the project. A public demo or source repository is not available here.