Atul Lal
Projects
Operator placement on edge systems in stream processing
Utilized Raspberry Pi devices and Apache Flink to optimize edge offloading for stream processing systems, reducing latency and data traffic while maximizing resource efficiency in real-time data processing.
Scalable Distributed MD5 Hash Matching
A project that designs a distributed system for cracking 5-character passwords using MD5 hash matching. The system employs a web interface, a management service, and multiple worker nodes to enable scalable brute-force attacks. Users can add or remove worker nodes, improving efficiency and adaptability.
Exploring the Home Mortgage Market using Spark on Google Cloud
Predicting loan approval factors for home mortgages using the 2016-17 HMDA dataset. Identifying key influences such as gender, race, income, and property type, lenders can make data-driven decisions, and enhance profitability while assessing credit risk accurately.
Video Compression using Deep Learning
Built a DNN based model which compressed video files using autoencoders.Compressed video matrix using vector decomposition, motion estimation and DNN based autoencoders.
Multithreaded Borůvka's algorithm
Worked on solving minimum spanning tree of a graph using Boruvka’s algorithm in a multi-threaded environment. Tested the multi-threaded application on multiple high performance 200-core super computers to find bottlenecks and compare performance.
Tiny face detection in the wild
The project targets tiny face detection using a GAN to enhance low-resolution images, boosting detection accuracy on thermal images. Implemented with Keras, a deep learning framework.