ML Researcher · MAY 2025 – APR 2026
Fine-Tuning Portfolio
PEFT of language models across text, vision, and reasoning domains.
Each notebook includes detailed explanations covering: model architecture, LoRA/QLoRA implementation, memory optimization strategies, training dynamics, and inference best practices.
Fine-Tuning Notebooks
- Qwen2.5-VL (7B) — Vision-language model fine-tuning for multimodal understanding
- DeepSeek R1 (8B) — Reasoning-optimized model with chain-of-thought fine-tuning
- LLaMA 3.2 (3B) — General-purpose instruction following on FineTome-100k
