Hi, I'm Michael Li.
A
Curious, quick-learning, and passionate researcher who enjoys crafting creative solutions to complex real-world problems using cutting-edge technologies.
About
I am a graduate computer science student at Carnegie Mellon University, studying for a Masters of Intelligent Information Systems. My main interests are machine learning and cloud systems: specifically, interactive data analysis and reinforcement learning, as well as architecting and engineering big-data cloud systems at scale.
- General Skills: Machine Learning, Cloud Computing, Full-Stack Development
- Languages: Java, Python, JavaScript, C++, C, HTML/CSS
- Data Science: PyTorch, NumPy, Scikit-Learn, SciPy, Matplotlib, Pandas
- Frameworks: Spring Boot, Express.js, Flask, Next.js, Ionic Capacitor, React.js, Vue.js, jQuery, EJS, Bootstrap, Material UI, Junit, Jest, Gatling, Cypress
- Services: Azure Kubernetes Service, HashiCorp Consul, Redis, NGINX, Apache Kafka, Keycloak, HashiCorp Terraform, Helm, GitHub Actions
- Databases: MongoDB, PostgreSQL, H2
I'm always looking for opportunities to collaborate and innovate. Let's get in touch and build something amazing!
Experience
- Worked with Fulfillment by Amazon (FBA) Reimbursement team
- Designed, implemented, and deployed a full-stack internal chatbot to answer reimbursement related questions with an internal knowledge base, using retrieval-augmented generation (RAG)
- Engineered prompts to optimize retrieval and generation, aligning with Helpful Honest Harmless (HHH) principle
- Skills: AWS, Bedrock, Java, TypeScript, Kendra Knowledge Base, Prompt Engineering
- Conducted research in the Social Reinforcement Learning lab, advised by Prof. Natasha Jaques
- Led project on improving neural combinatorial solver robustness to different distributions on the traveling salesman problem, published at NeurIPS MATH-AI
- Worked on adversarial robotic manipulation project, funded by Amazon grant
- Skills: Machine Learning, Robotics, Python, PyTorch, NumPy, Matplotlib, Isaac Lab
- Developed an innovative study plan system, using bandits-based algorithms to optimize educational outcomes
- Engineered an adaptive reading recommendation system which personalizes readings based on difficulty and topics, using aggregate user data tracked in up to 8 different ways
- Exposed 3 groups of Flask RESTful APIs for seamless communication between microservices
- Skills: Machine Learning, Python, PyTorch, NumPy, Matplotlib, Flask, Pandas
- Worked with around 150 students in the course, teaching core machine learning concepts such as regression, classification, and clustering
- Taught weekly quiz section for assigned group of 20 students to review and practice
- Held weekly office hours to clarify concepts and answer questions
- Skills: Machine Learning, Python, PyTorch, NumPy, Matplotlib, Pandas
Projects
Designed a genetic curriculum for improving TSP model robustness on distributions of practical interest.
- Proposed TSPLib50 dataset for measuring performance on "realistic" distributions
- Improved robustness of neural models on the travelling salesman problem
- Presented at NeurIPS MATH-AI 2024
A free, adaptive, open-source natural science (physics / bio / chem) trainer for motivated secondary school students.
- Adaptively recommend questions appropriate for each students' level
- Database of 4000+ questions, differentiated by 400+ unique tags
- Live at mutorials.org
Education
Degree: M.S. in Intelligent Information Systems
GPA: N/A
Degree: B.S. in Computer Science
GPA: 3.96/4.0
Relevant Coursework: Reinforcement Learning (579), Social Reinforcement Learning (599J), Interactive Learning (541), Deep Learning (493G), Machine Learning (446), Distributed Systems (452), Modern Algorithms (422), Computer Vision (455)