Shreya Therupally

• Contributing to Moohan, a next-gen XR head-mounted device (HMD) developed by Samsung, Google, and Qualcomm, featuring 14 cameras.
• Developing Android framework components in C++, with deep integration across AOSP, Camera HAL, and Linux-based system layers.
• Implemented IPC mechanisms including sockets, AIDL, and shared memory for inter-process communication across framework modules.
• Optimized camera pipeline by enabling partial frame delivery, reducing sensor-to-display latency by ~50%.
• Worked with OpenCV for real-time preprocessing, reducing latency in various camera use-cases.
• Boosted RGB camera FPS from 10 to 80 through performance tuning.
• Wrote unit tests using GTest to enhance code reliability and test coverage.
• Trained and mentored junior engineers, providing guidance and direction.

AR Zone :

• Maintained and enhanced AR Zone, Samsung’s centralized AR platform app integrating AR Emoji, Deco Pic, and AR Canvas.
• Ensured platform compatibility across One UI (T/U OS) versions and device types, including fold/flip models.
• Managed feature integration, permissions, and UX consistency across modules.

AR Emoji :

• Improved avatar rendering and animation in modes like Scene, Mask, Mirror, and Play.
• Resolved issues in face tracking and real-time expression mapping.
• Collaborated with camera and UX teams to align avatar behavior with system APIs.

DecoPic :

• Built and maintained real-time AR camera effects (live stickers, GIFs, overlays) using Android SDK with MVP/MVC.
• Led transition of DECO PIC to a standalone app in One UI 7, adapting app lifecycle and permissions model.
• Fixed crashes, memory leaks, and performance issues across flagship and mid-tier Galaxy phones.
• Managed app rollout across devices using Perforce and internal build systems.

AR Canvas :

• Updated features and visual elements enabling 3D drawing over camera feed.
• Stabilized GIF module, improved user experience and UI responsiveness.
• Delivered UI updates using Java/XML in line with Samsung’s AR design guidelines.

• Developed TensorFlow Lite parser operations for SNAP, Samsung’s on-device AI acceleration platform.
• Benchmarked and optimized model performance using custom single-op TFLite models.
• Designed and executed test cases for the Caffe parser within SNAP’s OnDevice framework.
• Contributed to Face3D, an Android app for facial landmark detection and 3D face modeling.
• Worked on AI-DHA, processing triangular mesh data for 3D human models, including cleaning, visualization, and optimization.

• Analyzed and resolved memory leaks and runtime crashes in the SmartThings Android application using MonkeyRunner and Memory Analyzer Tool (MAT).
• Improved application performance and stability through manual code optimization and legacy Java code cleanup.
• Gained hands-on experience in debugging large-scale Android applications within a commercial product environment.
• Participated in workshops and presentations related to projects to gain knowledge.
• Prepared project presentations and reports to assist senior staff.