Kedar J.

WORK HISTORY

EDUCATION

ABOUT KEDAR J.

I am excited about shipping production grade autonomy platforms where ML meets embedded systems - and where reliability, runtime, system safety matter a lot. I work on runtime ML infrastructure, software integration for perception sensors and system level integration for powertrain, actuators, safety systems. I have experience working with Linux network tuning, low-level embedded drivers development, sensor characterization, software/electrical integration, and perception runtime interfaces for spinning lidars, fmcw lidars, imaging radars, GNSS/IMU sensors. Also I graduated recently with a master\'s degree in computer science with a specialization in computing systems. I am working on some interesting side projects. The most active ones are:1. Concurrent queue (SPSC) implementation - benchmarking throughput performance of various queue implementations and exploring how cache, memory, CPU affects the performance https://github.com/kedar2017/Concurrent_Queue2. Custom memory allocator - started with a linear allocator to benchmark against other allocators. Other than latency performance, I am exploring how a custom allocator could prove useful in terms of tracking metrics, and reducing a lot of overheads for some traditional workloads like tiled GEMM, graph building, multi-threaded allocationhttps://github.com/kedar2017/MemoryAllocator3. High performance Bit Packing - exploring challenges involved in making a very high speed byte packing function which has countless applications. The most straightforward implementation has both compute-bound and memory-bound limitations. https://github.com/kedar2017/BitPacking4. https://github.com/steenl/PortAlchemy - More details on this coming soon. Contact info: