My GitHub Projects: A Complete Guide to My Open-Source Work and Repositories

By Rudra Sarker • Published March 20, 2026

Introduction

I am Rudra Sarker, a student developer and researcher from Shahjalal University of Science and Technology (SUST) in Sylhet, Bangladesh. Over the past few years I have built, shipped, and open-sourced a range of software projects spanning assistive technology, STEM education, mental health, and web development. Every significant project I have completed lives on GitHub, publicly accessible to anyone who wants to study the code, fork it, contribute to it, or build on it.

This post is a comprehensive guide to my GitHub presence — what repositories I maintain, how each one works, the technical choices behind each project, and how you can contribute. Whether you are a developer looking for an interesting open-source project to contribute to, a student curious about how these things are built, or a potential collaborator assessing my work, this is the post to read.

I believe deeply that sharing code publicly makes software better, makes developers better, and makes the world better. Open-sourcing forces you to write code as if someone else will read it — because they will. It is the best code review practice I know.

My GitHub Profile

My GitHub profile is at github.com/rudra496. The profile shows contribution activity going back several years, with regular commits across all active repositories. I try to maintain a consistent contribution cadence rather than bursting activity around deadlines, because consistent incremental progress is how complex software actually gets built.

My development philosophy, reflected across all repositories, can be summarised in three principles: build for real users with real needs; prioritise accessibility and performance from day one, not as afterthoughts; and document every significant decision so that future contributors (including future me) understand not just what the code does but why it does it that way.

Featured Repository: site (Portfolio)

The site repository is this portfolio website you are currently reading. The live version is at rudra496.github.io/site/. It is a fully static HTML/CSS/JavaScript site hosted on GitHub Pages — no build tool, no npm, no framework dependencies. Every file you see in the repository is served directly to the browser.

That "no framework" decision was deliberate. A personal portfolio site does not need React. The overhead of a JavaScript framework — larger bundles, more complex deployment, more attack surface — is not justified for a content-heavy site where server-side interactivity is not needed. The result is a site that loads in under one second on a fast connection, achieves Lighthouse scores above 95 across all categories, and will continue to work on any browser that renders HTML, indefinitely.

The SEO implementation is thorough: every page has a unique title and meta description, canonical URLs, Open Graph tags, Twitter Card tags, JSON-LD structured data for both content type and breadcrumb trail, a comprehensive XML sitemap, and a robots.txt that explicitly allows crawling. The result is strong organic visibility for relevant search terms.

Technically interesting aspects of the site include the 3D animated background on the portfolio home page (built with Three.js), a responsive navigation system that collapses to a hamburger menu on small screens, WebP images with JPEG fallbacks for all photography, and lazy-loaded Font Awesome icons (loaded via media="print" trick to avoid render-blocking CSS).

Contributions to the site repository are welcome, particularly for accessibility improvements, performance optimisation, and new blog post content.

Featured Repository: sightlineai

SightlineAI is my most technically ambitious project. The live demo is at rudra496.github.io/sightlineai/. SightlineAI is an AI-powered assistive eyewear system designed to help visually impaired users navigate their environment more safely and independently. The concept originated from a design challenge: how do you make computer vision practical in a low-cost wearable device that can run without internet connectivity?

The technical stack combines Python for the core computer vision pipeline with a Raspberry Pi as the embedded computing platform. The computer vision layer uses OpenCV for image capture and preprocessing, and a quantised MobileNet V2 model for object detection — the quantisation step (converting from 32-bit float to 8-bit integer weights) reduces the model size by approximately 75% while retaining most of its accuracy, making it practical to run inference at several frames per second on the Pi's ARM processor.

The user feedback system uses text-to-speech (via the pyttsx3 library for offline operation, or Google Cloud TTS for higher voice quality when online) to announce detected objects, their approximate distance (estimated from bounding box size relative to known object dimensions), and their position (left, centre, right) in the visual field. This spatial audio feedback gives users a mental map of their immediate environment.

Key files in the repository include detect.py (the main detection pipeline), models/ (quantised TFLite model files), audio/ (text-to-speech integration), and web/ (the GitHub Pages demo site that explains the system for sighted visitors). The README.md includes hardware assembly instructions, software setup steps, and a bill of materials so that anyone can build their own SightlineAI device for under $50 in component costs.

Contributions most needed: improving the distance estimation algorithm (currently based on a simple heuristic; a stereo camera approach would be far more accurate), adding support for Bengali and other South Asian language text-to-speech, and writing unit tests for the detection pipeline.

Featured Repository: science (3D Science Lab)

The science repository powers the 3D Science Lab platform. Like the portfolio site, it is a static site deployable on GitHub Pages — but internally it is much more complex, using Three.js and WebGL to render interactive 3D simulations in the browser.

The repository structure is organised by subject area: physics/, chemistry/, and biology/, each containing the HTML entry points and JavaScript modules for that subject's simulations. Shared utilities (OrbitControls, Raycaster, physics integrators, Three.js scene setup boilerplate) live in a lib/ directory to avoid duplication.

Adding a new simulation to 3D Science Lab is intentionally straightforward. The contribution guide explains how to create a new simulation module that exports a standard init(container) function and a dispose() function, and how to register it in the subject index. This architecture means contributors do not need to understand the full platform to add a single new experiment.

The platform is built with progressive enhancement in mind. If WebGL is not available (rare but possible on very old devices or in certain accessibility configurations), the page displays a static diagram of the simulation with descriptive alt text rather than a blank canvas.

MindWell and Other Contributions

MindWell — the open-source mental health platform available at mindwell-navy.vercel.app — is built with Next.js and TypeScript and deployed on Vercel. The development journey and architecture decisions behind MindWell are covered in extensive detail in my post Building MindWell: Lessons from Creating an Open-Source Mental Health Platform.

Beyond my own primary repositories, I contribute to other open-source projects in the web development and assistive technology ecosystems. I believe that even small contributions — fixing a typo in documentation, adding a missing ARIA label, submitting a bug report with a clear reproduction case — create real value in the open-source commons. A culture of small regular contributions to others' projects is as important to the open-source ecosystem as launching your own projects.

I also actively engage with the broader developer community through code reviews, issue discussions, and technical conversations in the repositories of projects I use. Good open-source citizenship means being a thoughtful, constructive participant in those conversations, not just a passive consumer of others' work.

My Development Philosophy

Several core principles guide how I approach every project:

Solve real problems for real people. Every project I have published was built to address a genuine need I observed — in my university, in my community, or in the broader world. SightlineAI came from thinking about how many visually impaired people in Bangladesh lack access to expensive commercial assistive devices. 3D Science Lab came from watching classmates struggle to visualise molecular geometry from a textbook. MindWell came from recognising how few free, privacy-respecting mental health resources exist for Bengali-speaking users.

Accessibility and performance are not optional. A tool that does not work for users with disabilities is incomplete. A tool that is too slow to use on a budget device excludes the users who most need it. Both are engineering problems with engineering solutions, and both deserve priority from the first line of code.

Documentation is code. A repository without good documentation is effectively closed-source in practice — the code may be public but it is not usable without understanding the context, conventions, and rationale behind it. Every repository I maintain includes a detailed README, inline code comments for non-obvious logic, and a CONTRIBUTING guide.

Measure, don't guess. Performance optimisations are only valid after profiling. UX improvements are only valid after user feedback. I try never to make a significant change based on intuition alone — measure first, then decide.

How to Contribute

All of my repositories welcome contributions. The general process is:

Fork the repository to your own GitHub account.
Clone your fork locally and create a new branch for your change (git checkout -b feature/my-improvement).
Make your changes, following the existing code style and conventions.
If the repository has tests, run them to verify nothing is broken (npm test, pytest, etc.).
Write a clear, descriptive commit message explaining what changed and why.
Push your branch and open a pull request against the main branch of the original repository.

Pull requests should be focused — one logical change per PR makes review much easier. If you are unsure whether a contribution is appropriate before investing time in it, open a GitHub Issue to discuss it first. Contributions I am always happy to receive: bug fixes, accessibility improvements, documentation improvements, new tests, and new features that align with the project's stated goals.

Other Profiles and Communities

My research profile on ResearchGate hosts my academic work and connects me with researchers working on related problems in assistive technology, EdTech, and health informatics. ResearchGate's recommendation system has introduced me to several research papers that directly influenced design decisions in my projects.

The United People Global forum is an international community focused on technology and social good. I participate in discussions there about the intersection of open-source software and community development, particularly as it relates to access to education and healthcare technology in the Global South.

Being part of these communities — GitHub, ResearchGate, United People Global — reinforces something I believe strongly: the most important unit in software development is not the individual developer or even the individual team, but the community of practitioners who share knowledge, challenge each other's assumptions, and collectively raise the quality of the work.

What's Coming

The roadmap for my GitHub presence includes several new repositories and significant updates to existing ones. In the near term:

SightlineAI v2: Stereo camera depth estimation, improved object tracking across frames, and a companion mobile app that can receive the detection output over Bluetooth for users who prefer a phone-based form factor over the wearable version.
3D Science Lab expansion: New simulation modules for optics, thermodynamics, and advanced biology, plus WebXR (VR) support for immersive laboratory experiences.
MindWell Bengali localisation: Full Bengali language support for both the UI and the condition guide content, making the platform accessible to tens of millions of Bengali speakers who currently cannot use it comfortably.
New project: SignTalk companion app: A mobile application companion to the SignTalk smart glove hardware, providing real-time sign language translation display and user customisation of the gesture vocabulary.

Follow my GitHub profile at github.com/rudra496 to stay updated on new releases and repository activity. Star the repositories that interest you to help others discover them, and feel free to open issues or discussions — I read and respond to all of them.

Connect With Me

Follow my work and connect across platforms:

Back to Blog