Experience
2022 — Now
2022 — Now
Santa Clara County, California, United States
• Lead end-to-end design and development of the AI Platform for Qualcomm's AI Gateway product — a heterogeneous compute platform executing AI workloads across NPU, GPU, and CPU for optimal performance and power efficiency.
• Architected and delivered an AI Applications Framework supporting modular deployment of LLM-based microservices on-device and in the cloud, with runtime scheduling across NPU/GPU/CPU backends.
• Drive LLM fine-tuning and deployment strategies using LoRA, QLoRA, and quantization tailored for Qualcomm's NPU/GPU-based edge silicon.
• Developed and maintained userspace drivers to interface AI workloads directly with Qualcomm NPU/GPU hardware, enabling low-latency, high-throughput inference without kernel-space overhead.
• Designed and deployed AI-powered network security applications: DNS Filtering, Anomaly Detection, and Intrusion Detection System (IDS).
• Led development of DPDK (PMD), eBPF, XDP, and AF_XDP-based userspace drivers for Qualcomm Networking and Wi-Fi chipsets, enabling kernel-bypass packet processing for high-performance networking.
• Designed userspace driver interfaces for direct hardware register access and DMA management on Qualcomm networking chipsets, reducing latency and CPU overhead.
• Managed performance optimization, profiling, and platform bring-up across diverse hardware configurations.
2014 — 2022
2014 — 2022
principal engineer and fastpath architect; led architecture, design, and development of a flexible DPDK-based dataplane platform for vBNG.
• Developed packet-processing application on Broadcom Jericho 2 programmable switch.
• Designed and implemented the full fastpath feature set: IP forwarding, GRE/GTP/L2TP/QinQ tunneling, L2/L3 ACLs, HQOS, CGNAT, HTTP Mirroring, Lawful Intercept, IP fragmentation, IP QoS, Parental Control, IoT fingerprinting, DNS web filtering, IPsec (Intel QAT + DPDK crypto).
• Ported platform to multi-cloud environments (AWS, Azure, VMware, GCP) for various core counts and memory configurations.
• Containerized deployments: CNF via Kubernetes + OpenShift + Multus + SRIOV; VNF via KVM, VMware ESXi, AWS, GCP.
• NIC support: Mellanox ConnectX-5/X6, Intel Columbiaville, and Fortville. Performance optimization using cache and profiling tools.
Virtual Service Edge Platform (vWAG / vCPE) — SD-WAN & Carrier Wi-Fi
• Dataplane Architect for Benu's Virtual Service Edge platform — combining Mobility, Wi-Fi, vCPE, and SD-WAN for cloud-managed residential and enterprise services.
• Led DPDK/VPP-based flexible dataplane platform supporting 2–56 cores, 2 GB – 512 GB memory, ConnectX-5 and XL 710 NICs.
• Fastpath features: IP forwarding, GRE/GTP/L2TPv3, L2/L3 ACLs, CGNAT, HTTP Mirroring, Lawful Intercept, IP fragmentation/QoS, Parental Control, IoT fingerprinting, DNS filtering, IPsec (Intel QAT + DPDK crypto).
• Platform deployed as VNF (KVM, VMware ESXi, AWS, GCP) and CNF (Kubernetes + Multus + SRIOV). Engaged directly with customer SEs to resolve field issues.
• Led NPU team developing fastpath and control-plane modules for a Wireless Access Gateway (used by Comcast and other ISPs) on Marvel HX 336B NPU and OCTEON.
• Designed and developed: ACLs, L2/L3 forwarding, Policing, ECMP, NAT, GRE, GTP tunneling, IPv6, IP QoS modules.
• Developed table management and control-plane software for NAT, GRE, GTP, ACL, and IP forwarding.
2011 — 2014
Greater Boston
• Led NPU team developing fastpath and control-plane modules for a Wireless Access Gateway (used by Comcast and other ISPs) on Marvel HX 336B NPU and OCTEON.
• Designed and developed: ACLs, L2/L3 forwarding, Policing, ECMP, NAT, GRE, GTP tunneling, IPv6, IP QoS modules.
• Developed table management and control-plane software for NAT, GRE, GTP, ACL, and IP forwarding.
2009 — 2011
2009 — 2011
Bangalore,India
Data Plane Technical lead. LTE - EnodeB Transport Team
• Developed fastpath modules on Cavium OCTEON 58xx and Intel Network Processor: IPSec, VLAN traffic differentiation, QoS, Ethernet OAM.
2007 — 2009
Designed and developed: Link & Board Redundancy, Static SA Redundancy for IPsec, NAT, IPv6, ACL (TCAM-based), Port Mirroring modules.
• SBC: developed data plane modules for IP QoS (DiffServ, DSCP Classifier, 6-tuple classifier, DSCP Marker, WRED) and RTP security.• Developed data plane modules for SIP Parsing, IP Reassembly, IP Fragmentation, NAT, ACL, Link & Board Redundancy for a carrier-grade IMS/Softswitch load balancer.
Education
Walsh College
Doctor of Philosophy - PhD
Walsh College
Master's degree
Texas McCombs School of Business
Postgraduate Degree
I2IT Pune
Master’s Degree
Jawaharlal Nehru Technological University
Bachelor’s Degree
bvk