Converting pcap expressions to BPF (Berkeley Packet Filters)

- Developed a compiler that translates pcap expressions into BPF, allowing for more efficient and accurate filtering of network packets
- Implemented improvements relative to tcpdump's capability to capture transport layer on top of ipv6
- Designed and implemented control flow graph optimization algorithms to improve BPF's execution speed and resource utilization
- Conducted extensive testing and debugging to ensure correctness and robustness
- Contributed to the open source community, enhancing readability of 'Caper expansion' section in
(, code is deployed online at

Web Server for BPF simulations

- Developed a user-friendly web server application in Go to facilitate packet analysis through user-uploaded pcap files and pcap expressions
- Generated graphical representations depicting the step-by-step packet processing flow based on applied BPF codes.
- Displayed raw packet data alongside corresponding BPF expressions to illustrate filtering logic
- Integrated a custom compiler that translated user-provided pcap expressions into valid BPF codes for seamless packet analysis
- Leveraged Go's concurrency features to efficiently handle multiple requests simultaneously
(Accessible at

Firewall on DPDK(Data Plane Development Kit) for High-bandwidth Network

- Developed a firewall solution utilizing DPDK to handle high-speed network traffic at 100Gbps bandwidth
- Conducted performance optimization through utilizing RSS load balancing and DPDK libraries
- Administered extensive testing and benchmarking to validate the program's stability, scalability, and performance

HULA (Hop-by-hop Utilization-aware Load balancing Architecture) Load balancer for Data Center

- Built a scalable, adaptable, and programmable congestion aware load balancer for programmable data plane using P4 language
- Developed an automated tool to generate fat-tree network topologies based on specified requirements and constraints
- Conducted extensive testing and validation to ensure the accuracy and reliability of the load balancer
- Constructed an automated visualizer to capture the path of packets across the network infrastructure


Paddy Disease Classification

- Engineered a model on a large-scale dataset, achieving an accuracy of 90% using Convolutional Neural Network (CNN) with Tensorflow
- Conducted thorough data analysis and data preprocessing to prepare the data for modeling
- Fine-tuned the hyperparameters to optimize model performance