Robotics data and Python/ML engineer based in Bengaluru. I build high-throughput sensor data pipelines, automated multimodal data QA engines, and dataset tooling for robot learning.
- RigSim — Multi-camera and IMU sensor rig simulator with physics-based fault injection, automated QA verification, and Foxglove MCAP streaming (verified 99/99 passing tests, 84% coverage, 100% data yield on 6-camera rig).
- EpisodeForge — Enterprise ETL and transformation engine converting multi-modal teleoperation episodes into benchmark-ready LeRobot, RLDS, Robomimic, and WebDataset releases (verified 54/54 passing tests, 66% coverage, 5 client specs validated).
- CodeAtlas — High-performance C++20/Qt6 code visualizer extracting compiler-grade AST knowledge graphs via libclang and local vector RAG (verified 1.00 ms AST extraction, 7/7 passing tests).
- OmniCore — Modern C++ computational foundations monorepo delivering zero-overhead primitives for computer vision, spatial geometry, and autograd (verified 589 ns KD-tree queries, 108/108 passing tests across 5 modules).
- Languages & Systems: Python, C++ (C++17/C++20), Bash, Linux
- Data & Robotics Tooling: PyTorch-adjacent tooling, MuJoCo, ROS2, OpenCV, DuckDB, PyArrow, PyAV, MCAP
- Engineering & Infrastructure: Docker, uv, CMake, GitHub Actions, GoogleTest, Google Benchmark
- Email: snshyam747@gmail.com
- LinkedIn: linkedin.com/in/shyam-sn
- GitHub: @Shyam-SN