Our Teams
Client Projects
Our client teams allow those with more experience the chance to work on real, industry-facing projects and build on their software development foundations.
What you'll do:
- Master new tech stacks by working on projects, such as data science, full-stack development, and back-end infrastructure.
- Work side by side with engineers at top companies, learning industry workflows and practices.
- Get a chance to receive feedback and constantly iterate on your work.
Mentored Project
Our mentored team provides the opportunity for those with no prior software engineering experience to learn the skills needed to take on industry-facing projects.
What you'll do:
- Learn good coding practices and build your own personal website from scratch.
- Design, develop, and deliver a full-stack web application for a nonprofit client.
- Learn modern frameworks and technologies like React, Node, and GraphQL.
SPRING 2025 PROJECTS
Client Projects
Our client teams work with industry partners to build products ranging from full stack web development to machine learning.
Amazon partnered with Codebase to upgrade DeltaCAT 2.0 by strengthening A.C.I.D. transactions, improving persistence, and adding Rivulet-convertible dataset support. These enhancements make the open-source data lakehouse more powerful for AI/ML-ready, industry-scale workflows.
Recidiviz partnered with Codebase to build an LLM-powered platform that uses automation to streamline parole and probation workflows and reduce bottlenecks, boosting efficiency for case managers and improving outcomes for formerly incarcerated individuals.
Sony PlayStation partnered with Codebase to develop a BERT-based NLP model that analyzes chat threads in real time to flag high-risk behaviors. Built with Python and PyTorch, the scalable solution enables early intervention and reinforces PlayStation’s commitment to player safety.
Sony PlayStation partnered with Codebase to create a PyTorch-based ML classifier that identifies video game screenshots versus real images. The solution reduces reliance on third-party moderation, lowers costs, and improves player safety with a scalable in-house approach.
Mentored Project
Our mentored team focuses on learning the essentials of software development and simultaneously develops a full-stack web application for a non-profit organization.
Lotus Bloom, a community organization serving 1,000+ families, partnered with Codebase to build a secure online forum for staying connected during school breaks. The custom platform supports resource sharing, strengthens family bonds, and advances Lotus Bloom’s mission of community empowerment.
Development Timeline
Here’s a breakdown of how our projects are run every semester.
![]()
Before the Semester
Over break, our leadership team works together to source and scope new projects with industry clients and non-profits.
![]()
Project Kickoff
After welcoming our new members and a club-wide retreat, we form project teams based on developer preferences! Everybody is onboarded onto their projects’ tech stacks.
![]()
Mid-Semester Deliverable
Around the halfway point, our teams visit our clients and present a deliverable of what they’ve accomplished so far. We demo what we’ve built, take constructive feedback, and go over next steps.
![]()
Final Deliverable
At the end of the semester, we present a polished version of our product, go over project features, and give our clients a chance to ask questions. After, we celebrate a semester’s hard work at banquet!
![]()
Technical Handoff
In the final week, our developers and PMs coordinate with our clients to clarify documentation and assist in the integration of our code into their codebase.
Past Projects
Every semester we take on five new projects with high growth tech companies. Take a look at some of our past projects!
Amazon’s big data team partnered with Codebase to optimize dataset merging in Apache Spark, reducing vCPU usage and improving efficiency at scale. Using Java/Scala on AWS infrastructure, Codebase implemented advanced merging strategies that accelerate petabyte-scale processing, lowering costs and powering ML pipelines for recommender systems, fraud detection, and other large-scale AI applications.
Atlassian provides an arsenal of productivity tools, including Trello, Jira, and Bitbucket, that powers 125,000 companies worldwide. We built an analyzer for Atlassian's decision logs on their authorization and authentication policies.
DoorDash is an online food ordering and food delivery platform. Our team built a production-like testing platform from 0 to 1 to rigorously evaluate and test ML Engineers’ pytorch models. This deployed environment allowed faster model development and validation, as well as provided predictive insights to prevent issues from occurring in production systems.
Etsy is an e-commerce platform that allows users to sell and purchase a wide variety of handcrafted and vintage items. For our project, our team accomplished a proof-of-concept of certain portions of Etsy's Image Upload pipeline in order to better understand how a future, re-architected efficient/optimized pipeline might be built on Etsy's new service platform.
Google Labs is Google's home for the latest AI experiments and technology. Codebase developed Sparky, a mobile app for Google's AI Studio, which allows developers to prototype generative AI models. Sparky makes it easier for developers to test multimodal prompts and interact with Google's Gemini API.
Meta builds technologies that help people connect, find communities and grow businesses. Meta's video team wanted to optimize brute-force algorithms for video compression, a crucial part of video processing on any device or application. Codebase researched and developed statistics-based algorithms and machine learning models to enhance video compression solutions and optimize compute power while preserving video quality.
Meta builds technologies that help people connect, find communities and grow businesses. For our second Meta project, Codebase worked on the Facebook General Matrix Multiplication (FBGEMM) Pytorch library to extend auto-vectorization of matrix multiplication operations for ARM CPUs, enabling Meta's recommendation models to run efficiently on ARM architecture.
Nuro exists to better everyday life through robotics, and its custom electric autonomous vehicles provide a convenient, eco-friendly alternative to driving, making streets safer and cities more livable. Codebase will be developing a kiosk that is integrated with badge reader hardware to allow operators to efficiently clock in & out according to their assignments for the day.
PayPal’s engineering productivity team partnered with Codebase to build AI-driven workflows that streamline software delivery and coordinate across critical tools and services. The system enhances collaboration, improves efficiency, and drives meaningful impact directly within PayPal’s production development processes.
Sourcegraph allows developers to rapidly search, write, and understand code by bringing insights from their entire codebase right into the editor. Our team built a machine learning training and evaluation pipeline for Cody, Sourcegraph's AI-powered code autocomplete assistant, by ranking and integrating varied context sources like Issue Tickets and Documentation, as well as evaluating its effects on LLM performance.