Principal Software Engineer
CurrentAI/ML Training Infrastructure Solution Development • Design and developed multiple components in AI-ML workload generation solution, including the controller and the engine with RoCEv2 and TCP protocol stack.• Designed and developed a data flow engine for AI/ML workload generation with NICs from Broadcom and Mellanox.• Implemented QP-negotiation capabilities and QP-REUSE support in the data flow engine.• Achieved interoperability with customer DUTs (Meta) using different collective communication methods.• Integrated third-party stack NCCL into Keysight’s AI/ML solution.• Added flexibility and accurately measured data center infrastructure performance. Wireless (5G-UE) ORAN MU-MIMO Solution (TSA): • Developed a middleware framework to integrate high-scale hardware for ORAN traffic generation on a Kubernetes platform.• Developed middleware microservice framework for high-scale ORAN MU-MIMO solutions for eCPRI traffic.• Created driver microservice to parse and extract flow and C/U plane traffic, programming the FPGA for multi-port and multi-card operations.• Designed and integrated a GRPC-based M-Plane app with the middleware server.• Developed REST API-based automation regression test infrastructure to maintain solution quality.• Created a software framework to distribute thousands of UEs and associated traffic across multiple virtual nodes.• Enabled Keysight’s wireless solution on AWS-cloud platform, providing emulated router support.• TWAMP Protocol Stack Development to improved RTT and latency calculation for 5G infrastructure components.• Integration of new platform (eLSU) in Keysight’s wireless product solution. • Ensured ixload traffic engine could pump L4-7 traffic to ELSU's PDCP.• Multi-VM Infrastructure for 5G Ecosystem :Architected and developed support for ixload traffic generation, enabling higher throughput traffic simulation across different VMs.