🔭 LinkedIn: linkedin.com/in/siddd27
📫 Email: siddharth.rambhia@gmail.com
😄 Pronouns: he/him
🧩 CodeForces: CodeForces
I’m an active contributor to the Kornia ecosystem, focusing on high-performance computer vision across both Python (PyTorch) and Rust. My work generally centers on removing bottlenecks, optimizing memory layouts, and making complex vision tasks viable for resource-constrained environments.
- Memory-Efficient AprilTags: Reworked the core data structures for AprilTag detection in Rust. By optimizing how data is stored, I reduced memory consumption from >6GB to ~450KB (a 14,000× reduction), enabling the library to run on small edge devices.
- High-Speed Retrieval (kornia-bow): Authored a pure-Rust Bag of Words (BoW) crate. Through careful memory management, it achieves 11× faster lookups and 5× lower memory usage than reference implementations.
- Massive Python Speedups: Optimized several core PyTorch operations, including:
- Otsu Thresholding: Achieved a 308× performance increase through strategy refinement.
- Geometric Estimation: Implemented vectorized analytical polynomial solvers, resulting in a 200× speedup.
- 3D Masking: Refactored
bbox_to_mask3dfor a 55× improvement.
- Pipeline & Algorithm Tuning: Developed Fused Preprocessing Kernels to minimize memory bandwidth, delivering a 5× speedup in image preprocessing pipelines.
- Optimized the Needleman-Wunsch algorithm for better cache efficiency (17.7× speedup).
- Refined SVD (Singular Value Decomposition) implementations to maximize throughput for real-time vision tasks.
all this and a lot moreeeee! I'm always up to help optimize real world projects, if you have a intresting optimization problem, feel free to share and we could tackle it together!
| Languages |
|
| Frameworks & Libraries |
|
| Databases & Cloud |
|
| Tools |
|
👋 About Me
- Third-year Computer Science student
- I love to random trivia and quizzing!
- Love making fast things even faster!
- Come chat if you’ve already made it this far :)
🏆 Achievements
- Codeforces Rated 1810
- Hackathons: Winner of the codeflash optimization hackathon
- Community: Convener of the AI club at DAU (formerly DAIICT)