lucyd is a music recommendation tool developed by four graduate students in UC Berkeley's Master of Information and Data Science (MIDS) program.
lucyd let's you ask for music recommendations using whichever terms you want. lucyd seeks to demystify the typical "black box" music recommendation algorithm by not only giving you the freedom to customize your recommended playlists, but by giving you interactive tools and visualizations so you learn about your music while you explore.
lucyd offers something different. Our algorithm focuses on you--how do you perceive a song, and how can that be used to find new ones?
lucyd is a tool for people who are curious about music. If your favorite pasttime is perusing music blogs, clicking into the depths of your streaming service, or curating your own playlists, lucyd can provide you with more tools to navigate the worlds music.
Designing lucyd required several steps:
- Algorithms: Outline framework and requirements of a music recommendation algorithm based upon user-determined "tags", rather than song and/or artist attributes.
- Data Source(s): Identify potential data sources to "teach" the recommendation algorithm.
- Visualizations: Develop visualizations that help the user understand the songs in their playlist, and the ones nearby
- Application Framework: Combine all of these elements into a live web application that allows the user to influence and learn from the songs recommended by lucyd.
Product Design & Web Development
github | LinkedIn
Tim is an economic consultant in Boston, MA where he works on complex data modeling engagements. For lucyd he was responsible for the product design and connecting the work of the team into a web application. In his free time he enjoys long walks on the beach. He also enjoys short walks on the beach.
Primary Skills Developed: Django web framework (+ HTML, CSS, JavaScript), Cloud serverless web development (AWS: API Gateway, Lambda, IAM, CloudWatch, CloudFront, Route 53, Certificate Manager)
Cloud Data Architect
github | LinkedIn
Kyle is a federal data science consultant in Washington DC. His professional work focuses on interacting with Medicare health claims databases and designing auditing and monitoring algorithms to identify non-compliance or risk. He leveraged this experience to scale lucyd's cloud infrastructure, developing the databases, data pipeline, and query structure that supports user interaction with the data.
Primary Skills Developed: Cloud Architecture (Amazon RDS mySQL, DynamoDB, DataPipeline, S3); Cloud Machine Learning (AWS EC2, AWS Sagemaker); Cloud serverless web development (AWS API Gateway, AWS Lambda, AWS IAM User Management); Python; SciPy
Algorithmic Design
github | LinkedIn
Mike is a Chicago-based derivatives trader who enjoys live music and his mini Bernedoodle, Hugo. At a young age he developed an internal algorithm to identify the most efficient checkout line at any given supermarket; rumor has it he has never been wrong. He tried to replicate some of this magic in the development of lucyd, where he focused on algorithmic design and user interactivity.
Primary Skills Developed: Python, AWS cloud machine learning & serverless web development (API Gateway, Lambda, SageMaker, S3)
User Interface & Project Management
github | LinkedIn
Hanna is a technology consultant working in the Customer Experience (CX) industry out of Bozeman, Montana. For most of the project, Hanna was responsible for building visualizations in D3, but in the final week discovered AJAX and re-arranged how everything on the front-end fit together so she could add some progress spinners. Thankfully, it worked out.
Primary Skills Developed: HTML, CSS, JQuery, JavaScript & D3
We developed lucyd during the course of our Capstone semester in UC Berkeley's Master of Information and Data Science (MIDS) program. MIDS offers a comprehensive curriculum in data science, including courses focused on computer science, machine learning at scale and legal, policy and ethical issues.