Hi, I'm Sagar
Graphics engineer, creator and teacher. The person behind rgbguy.
Graphics engineer. Nine years of building experiences that run at high framerates in the real world.
I started as a graphics and game developer, then moved to XR research at Qualcomm and filed patents. After that came native rendering pipelines, camera SDKs and FaceAR products, scaled to millions of users across Android and Web. Not just shipping features. Architecting the systems underneath them.
Diving deep into graphics has always been the dream. So in 2024 I went back to first principles. I rebuilt concepts from the ground up and implemented rendering algorithms on multiple graphics APIs and platforms. Then I turned the learnings into videos that 300K+ learners watch.
Right now I'm building a 3D engine from scratch in C/C++. Because the best way to understand a system is to build it yourself.
The rare kind of engineer who reads papers, ships production code, and teaches it to a room full of people. Sometimes all in the same week.
- Dedicated myself to deeply studying computer graphics and real-time rendering, building custom rendering systems and implementing core graphics algorithms from scratch. All work documented in the lab. Worked under the mentorship of Dr. Vijay D. Gokhale, consistently ranking at the top of the batch of 200+ students at Astromedicomp.
- Grew RGBGUY from 800 followers to 410K+ across platforms (Instagram 245K+, Facebook 90K+, YouTube 75K+) and 10M+ views on YouTube, reaching millions of learners every month.
Rebuilt a WebGL Virtual Try-On SDK from scratch. Performance went from 10 FPS to 30+ on low-end devices and 60+ on high-end. Flexible architecture for quick SKU creation and deployment. Led client launches including a high-traffic Zepto x Holi collab.
Owned development and full release cycle of Shutter, a lightweight cross-platform C++ graphics SDK built on OpenGL and Metal. It powers ShareChat's Motion Video, Video Editor, and Camera. Built the FaceAR filter pipeline, Face3D overlay, Facemask, FaceMorph, and Image/Video overlays, along with tooling for designers to author effects directly.
Novel real-time depth compression algorithm for cloud VR reprojection on a phone-HMD split architecture, reducing judder and improving visual quality. Drove the full pipeline from literature survey to implementation and validation.
Synthetic Data Generation: Built a pipeline in Blender + Python producing 15+ TB dataset for training for a temporal/spatial light estimation model. Camera pose, light settings, and scene params in; stereo and panoramic images with depth buffer out.
Where it started. Built multiplayer VR simulations covering VR interactions, custom components, inverse kinematics, game AI, UI and audio systems, networking, and packaging.