Jun 2020 – May 2024 · 3 yrs 11 mos · South Korea

• Developed highly optimized 5G wireless algorithms in C++17, leveraging SIMD to achieve superior performance to deploy on base stations of network operators like Verizon
• Managed a team of four developers in the design and implementation of the Channel Estimation Module
• Achieved 30% performance boost by optimizing algorithms with SIMD instructions, enhancing loop structure, improving memory access patterns, and exploiting instruction-level parallelism
• Led the optimization efforts, coordinating with Intel engineers to adapt the software for new Sapphire Rapids processors, which led to 70% performance improvement along with an additional 20% power savings
• Utilized advanced hardware-assisted profiling techniques, such as Last Branch Records (LBR) and Precise Event-Based Sampling (PBES), to analyze workloads and pinpoint performance bottlenecks
• Implemented comprehensive E2E performance tests to simulate real-world scenarios and optimised E2E unit test running time 6-fold

Sep 2018 – May 2020 · 1 yr 8 mos · South Korea

• Developed a kubernetes-based platform for Telco-grade carrier networks which can seamlessly use openstack networking
• Enhanced Kuryr-Kubernetes' scalability, removing resource leaks, handling exceptions, and contributing selected improvements to upstream for telco-grade networks
• Enabled AWS support by using GRE based overlay tunnels to emulate VLAN tags due to absence of native support

May 2018 – Sep 2018 · 4 mos · India · Remote

• Optimized the Linux GPU task scheduler which is used by GPU drivers including AMD, Vivante, and Broadcom
• Implemented dynamic load balancing of jobs based on current load replacing the previous static round-robin policy, by enabling shifting of jobs across hardware engines while handling arbitary dependencies across jobs
• Integrated the simplified API and bug fixes into GPU drivers which was shipped as part of linux kernel v4.20

May 2017 – Jul 2017 · 2 mos · Gyeonggi, South Korea

• Worked on the NFV MANO software stack which is part of Samsung's network virtualization solution
• Worked with various open-source technologies like OpenStack, OpenSource MANO, Docker, Kubernetes
• Contributed to Open Network Automation Platform (ONAP) and compared its features with Samsungs solution

May 2016 – Mar 2017 · 10 mos · Greater Lucknow Area

• Contributions: https://cgit.freedesktop.org/mesa/mesa/log/?qt=author&q=Nayan+Deshmukh
• Implemented interpolation algorithms to support a richer feature set of VDPAU spec which allows end-user
• applications like VLC media player access to video decompression acceleration hardware on the GPU
• Implemented the DRI3 protocol for userspace GPU drivers for laptops with discrete GPUs to enable allocation of frame buffers by clients thereby removing the expensive copy operations
• Utilized the newer API in VDPAU implementation thereby improving the latency and reducing memory bandwidth