Machine Learning
AUgarde · Jan 2025 – Jun 2025
- Built pipelines to collect, clean, deduplicate and normalize conversational datasets (JSON/JSONL) for supervised fine-tuning (SFT).
- Prepared and validated training data for Unsloth-based fine-tuning, including tokenization and formatting checks.
- Added quality filters to remove low-quality samples [add: dataset size and % removed].

