Radhesh Harlalka
Development
California, United States
Skills
Data Science
About
Radhesh Harlalka's skills align with Programmers (Information and Communication Technology). Radhesh also has skills associated with IT R&D Professionals (Information and Communication Technology). Radhesh Harlalka appears to be an entry-level candidate, with 21 months of experience.
View more
Work Experience
Software Engineering Intern
Veritas Technologies LLC
May 2023 - August 2023
- • Developed unsupervised learning pipeline based on DBSCAN clustering for ransomware attack detection. • Used dedupe ratio and kilobytes transferred features to detect anomaly with statistical histogram comparison. • Significantly reduced false positives across Oracle and Vmware workloads when subjected to rigorous testing. • Deployed the model training and testing workflow over Azure cloud using Kubeflow for real-time detection.
Computer Vision Engineer
HyperVerge Technologies LLC
June 2021 - July 2022
- • Developed a face size invariant anti-spoofing system by deploying vision transformer on PyTorch framework. • Structured Swin Transformer in ensemble with CDCN++ network for robust detection of mask attacks. • Developed auxiliary supervised deep CNN model for identification of print and replay attacks accurately. • Attained SOTA f1-scores of 99.3, 98.9, 99.4 on CelebA spoof, CASIA and Rose datasets.
Machine Learning Engineering Intern
December 2020 - March 2021
- • Applied YOLO-v4 object detection algorithm to identify humans in live videos using custom data generator. • Modeled ConvLSTM neural network architecture on Tensorflow to capture temporal and spatial information. • Boosted accuracy from 83.4% on pretrained VGG16 to 92.71% on self devised custom ConvLSTM model.
Research Assistant
Indian Statistical Institute, Computer Vision Lab
May 2019 - July 2019
- • Created a novel architecture inspired from Pix-to-Pix GAN for reconstruction of characters in degraded text images. • Designed mask and predict strategy to artificially mask foreground text pixels. • Modeled custom U-Net Convolutional Neural Network to segregate text and background with higher accuracy. • Achieved an F-Measure and DRD of 96.89% and 1.53 respectively on eminent DIBCO 2018 dataset. Projects • Responsive Web Application Development: Created a web app to search for events using TicketMaster API, hosted on Google Cloud Platform. Used Node.js and Angular 8 for development. • Mobile Application Development: Developed an event finder mobile application on android studio using JAVA. Used Volley library to make fast synchronous HTTP requests for posting and fetching data from the backend. • Weenix OS: Developed a Unix based kernel. Built processes and synchronized threads using mutexes and condition variables and implemented a scheduler to allow context switching. Supported S5FS file system. • OMR Checker: Developed an OMR checker system which displayed the score using the image of OMR sheet. Used semantic segmentation which handled misaligned images and performed with 100% accuracy.
Education
University of Southern California
Masters of Science in Computer Science
August 2022 - May 2024
