Daily learnings, achievements, and project milestones
Day Streak
Projects Completed
Commits This Month
Hours Learned
Built a complete Vision Transformer from scratch using PyTorch. Implemented multi-head attention, patch embedding, and position encoding. Achieved 85% accuracy on CIFAR-10.
Studied Parameter-Efficient Fine-Tuning techniques. Implemented LoRA for fine-tuning LLaMA-7B on custom dataset with 10x memory reduction.
Developed an end-to-end object detection system using YOLOv8. Trained on custom dataset with 5000+ annotated images. Deployed with FastAPI backend.
Successfully completed Columbia University's DLCV course. Covered CNNs, object detection, semantic segmentation, GANs, and diffusion models.
Deep dive into diffusion probabilistic models. Implemented DDPM from scratch and fine-tuned Stable Diffusion on custom art dataset.
Built a Retrieval-Augmented Generation chatbot using LangChain, ChromaDB, and OpenAI API. Indexed 1000+ technical documents for domain-specific Q&A.
Implemented U-Net architecture for medical image segmentation. Trained on brain MRI dataset with Dice coefficient of 0.92.
Developed a multi-modal model combining text (BERT) and image (ResNet) features for social media sentiment analysis. Achieved 88% accuracy.
Published comprehensive guide on "Understanding Attention Mechanisms in Deep Learning" - received 500+ views in first week.
Learned experiment tracking, hyperparameter tuning, and model versioning. Set up automated ML pipeline with W&B sweeps.
Built stock price prediction model using LSTM networks. Implemented attention mechanism for improved long-term dependency capture.
Began Stanford's RL course. Implemented Q-learning and DQN for simple game environments (CartPole, Atari).