Biswajit Banerjee Email & Phone Number

Atlanta, Ga, Us

Atlanta, Georgia, United States

◦ Researching towards making the metabolic fold space searchable through creating protein fold level embedding and aligning them to their functions. ◦ Co-authored a research paper on RiboVision2, a novel web server employing Graph Machine Learning for comprehensive ribosomal RNA visualization, invited for publication in the Journal of Molecular Biology. ◦ Engineered a novel pipeline to map metabolic pathways to catalyzing protein folds, integrating multiple public databases and resolving ambiguities. Presented findings at the Exploration and Origins Colloquium 2024.

Seattle, Washington, United States

Developed a compact transformer-based generative language model using network security detection data across 2271 customer networks. This identified groups of related detection events that have a high likelihood of indicating cyber security kill-chain activity. It reduced around 68% of false positives and lowered user alert fatigue.

Bengaluru, Karnataka, India

◦ Developed an end to end open-set object detection system for cattle verification utilizing custom tuned MobileNetv2 architecture to accurately distinguish registered cattle (in-set) from unregistered cattle (out-of-set). ◦ Prioritized model optimization efforts using the 80/20 rule, putting priority on runtime optimization while time boxing accuracy improvement, achieving a 2ms forward pass while maintaining 84% accuracy for in-set and 92% for out-of-set cattle identification. ◦ Trained YOLOv6 network to identify the number of distinct farm animals present in the frame with an MaP of 88% at IoU 0.95 and deployed in mobile devices. ◦ Had the opportunity of mentoring interns.

Bangalore Urban, Karnataka, India

◦ Created algorithm to optimize physical circuit module placement, which was featured in Forbes. ◦ Engineered a modular pipeline to utilize spark and airflow to generate training data with dynamic transformations for different data requirements. This required cross-team collaboration and data moderation. ◦ Created a pipeline to train a slack prediction (regression) model which replaced 3 weeks long interpolation method to 2-3 hours of training and inference with margin of error of 10 microseconds. ◦ Identify which timing path needs optimization by casualty analysis which resulted in 45% less power consumption and 30% performance improvement for ARM series chip-sets. ◦ Created chip congestion detection by aggregated complex flow traffic data into intuitive heatmaps.

Master'S Degree, Bioinformatics

Bachelor Of Technology - Btech, Computer Science