Hey there! I'm Melody

About me:

• Currently pursuing a Master's degree in Software Engineering at Northeastern University.
• I serves as a Teaching Assistant of graduate students in Northeastern university from Sep 2025 until now.
• I'm on track for learning more about Artificial Intelligence, Systems Design, and Cloud Architecture.
• How to reach me: zhao.y4@northeastern.edu
• Fun fact: I've visited more than 56 countries.

Interests

• AI Software Development
• Machine Learning
• Web Development(Full Stack)

Blogs

• Personal website and blog

Software & AI Projects | Open Source Contributions

• Built AI-agent travel chatbot with Agno, Groq, and ChromaDB; fine-tuned model weights and open-sourced them.
• Open source contributor to notable open-source projects including:
  • HuggingFace Transformers
  • LangChain
  • JAX
• Studied the foundational paper "ZeRo" and implemented a DeepSpeed + PyTorch project for efficient distributed training and evaluation, optimizing resource usage and training speed for large-scale deep learning tasks.
• Implemented data parallelism for MNIST training with JAX's shardmap sharding API and contributed to OSS JAX.
• Built and extended Micrograd, a minimal autograd engine, as part of self-study of Karpathy's Zero to Hero series:
  • Implemented the Value class, topological sort–based backpropagation, and dynamically constructed computation graph (DAG).
  • Added custom operations (e.g., exponentiation, ReLU, tanh), visualized graph structures, and built training loop for an MLP.
  • Reinforced understanding of the chain rule, gradient descent, and autograd principles through end-to-end implementation.

Projects:

1. Travel-Agent-App

• AI project to plan travel itineraries.
• Built with Python and machine learning to provide personalized travel recommendations.

2. micrograd

• Neural network implementation in Python.
• Focuses on understanding how neural networks work by implementing backpropagation from scratch.

3. Weekly-DSA-Journey

• Learning by building. Learn by sharing.
• This repository contains weekly solutions to Data Structures and Algorithms (DSA) challenges.

4. huggingface/transformers

• The model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models for both inference and training.
• Contributed to the open-source community by improving the usability and performance of this powerful library.
• Python, PyTorch, TensorFlow

5. final-project-digital-marketing-solar-sisters

• Digital marketing project based on the Solar Sisters initiative to promote sustainable energy solutions.
• Python, Data Analysis, Marketing

6. Deep-Learning-Training-and-Evaluation-with-DeepSpeed-and-PyTorch

• Large-scale deep learning model training and evaluation with DeepSpeed and PyTorch.
• DeepSpeed, PyTorch, Machine Learning

7. transformers

• The model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models for both inference and training.
• Contributed to the open-source community by improving the usability and performance of this powerful library.
• Python, PyTorch, TensorFlow

---