Welcome to the site for BeagleBoard.org GSoC 2025 projects!

Ideas#

../_images/ideas-below.webp

How to participate?

Contributors are expected to go through the list of ideas dropdown below and join the discussion by clicking on the Discuss on forum button. All ideas have colorful badges for Complexity and Size for making the selection process easier for contributors. Anyone is welcome to introduce new ideas via the forum gsoc-ideas tag. Only ideas with sufficiently experienced mentor backing as deemed by the administrators will be added here.

Complexity

Size

High complexity

350 hours

Medium complexity

175 hours

Low complexity

90 hours

Tip

Below are the latest project ideas, you can also check our our Old GSoC Ideas and Past Projects for inspiration.

Deep Learning Medium complexity 175 hours

A Conversational AI Assistant for BeagleBoard using RAG and Fine-tuning

BeagleBoard currently lacks an AI-powered assistant to help users troubleshoot errors. This project aims to address that need while also streamlining the onboarding process for new contributors, enabling them to get started more quickly.

Goal: Develop a domain-specific chatbot for BeagleBoard using a combination of RAG and fine-tuning of an open-source LLM (like Llama 3, Mixtral, or Gemma). This chatbot will assist users with troubleshooting, provide information about BeagleBoard products, and streamline the onboarding process for new contributors.
Hardware Skills: Ability to test applications on BeagleBone AI-64/BeagleY-AI and optimize for performance using quantization techniques.
Software Skills: Python, RAG, Scraping techniques, Fine tuning LLMs, Gradio, Hugging Face Inference Endpoints, NLTK/spaCy, Git
Possible Mentors: Aryan Nanda

Linux kernel improvements Medium complexity 350 hours

Update beagle-tester for mainline testing

Utilize the beagle-tester application and Buildroot along with device-tree and udev symlink concepts within the OpenBeagle continuous integration server context to create a regression test suite for the Linux kernel and device-tree overlays on various Beagle computers.

Goal: Execution on Beagle test farm with over 30 mikroBUS boards testing all mikroBUS enabled cape interfaces (PWM, ADC, UART, I2C, SPI, GPIO and interrupt) performing weekly mainline Linux regression verification

Linux kernel improvements Medium complexity 175 hours

Upstream wpanusb and bcfserial

These are the drivers that are used to enable Linux to use a BeagleConnect Freedom as a SubGHz IEEE802.15.4 radio (gateway). They need to be part of upstream Linux to simplify on-going support. There are several gaps that are known before they are acceptable upstream.

Goal: Add functional gaps, submit upstream patches for these drivers and respond to feedback
Hardware Skills: wireless communications
Software Skills: C, Linux
Possible Mentors: Ayush Singh, Jason Kridner

Automation and industrial I/O Medium complexity 175 hours

librobotcontrol support for newer boards

Preliminary librobotcontrol support for BeagleBone AI, BeagleBone AI-64 and BeagleV-Fire has been drafted, but it needs to be cleaned up. We can also work on support for Raspberry Pi if UCSD releases their Hat for it.

Goal: Update librobotcontrol for Robotics Cape on BeagleBone AI, BeagleBone AI-64 and BeagleV-Fire
Hardware Skills: basic wiring, motors
Software Skills: C, Linux
Possible Mentors: Deepak Khatri, Jason Kridner

RTOS/microkernel imporvements Medium complexity 350 hours

Upstream Zephyr Support on BBAI-64 R5

Incorporating Zephyr RTOS support onto the Cortex-R5 cores of the TDA4VM SoC along with Linux operation on the A72 core. The objective is to harness the combined capabilities of both systems to support BeagleBone AI-64.

Goal: submit upstream patches to support BeagleBone AI-64 and respond to feedback
Hardware Skills: Familiarity with ARM Cortex R5
Software Skills: C, RTOS
Possible Mentors: Dhruva Gole, Nishanth Menon

Visit our forum to see newer ideas being discussed!