Pinned Loading

1. llm-evaluation-framework llm-evaluation-framework Public

   Research-grade LLM evaluation framework comparing Base Mistral 7B vs LoRA fine-tuned Medical AI — +17.8% ROUGE improvement, 65% win rate, interactive Plotly dashboard

   HTML

2. medical-qa-lora medical-qa-lora Public

   Fine-tuned Mistral 7B on 5,000 medical QA examples using LoRA + 4-bit QLoRA — 0.36% trainable parameters, val loss 0.6573, trained for $0 on Google Colab T4 GPU

   Jupyter Notebook

3. sentiment-analyzer sentiment-analyzer Public

   Fine-tuned DistilBERT on 25,000 IMDb reviews — 91.19% accuracy, beats TF-IDF baseline by +1.73%, live demo on HuggingFace Spaces, trained on M3 Mac CPU

   Jupyter Notebook

4. skin-disease-classifier skin-disease-classifier Public

   EfficientNetB0 trained on HAM10000 — 74.15% accuracy across 7 skin disease classes, Grad-CAM explainability, class-weighted loss for imbalanced medical data, Google Colab T4 GPU

   Jupyter Notebook

5. sports-rag-chatbot- sports-rag-chatbot- Public

   Full-stack AI sports analytics chatbot — RAG pipeline with LangChain + ChromaDB + Llama 3.2, FastAPI backend, custom Slack-style dashboard with dark/light mode, zero API cost

   Jupyter Notebook

6. yolo-object-detector yolo-object-detector Public

   Real-time object detection using YOLOv8 Large — 80 COCO classes, adjustable confidence, 4 model sizes, live demo on HuggingFace Spaces, detects people, vehicles, animals, food and more

   Jupyter Notebook