Karthik Pullalarevu

kpullala [at] andrew.cmu.edu

I am an M.S. in Computer Vision student at Carnegie Mellon University, in the Robotics Institute. My research interests span 3D vision, perception for autonomous driving, and multimodal AI.

Previously, I worked as a Lead Machine Learning Engineer - Computer Vision at HyperVerge, 150-member, bootstrapped and profitable AI startup operating across India, SEA, and Africa. I led R&D for the Face Fraud Detection team, developing and optimizing 10+ deep learning models for liveness and deepfake detection. These solutions now power 20+ million monthly identity verifications and contribute to 30% of the company’s revenue.

I completed my B.Tech in Electronics and Computer Engineering at Vellore Institute of Technology, where my research was focused on agritech and medical diagnosis using AI and computer vision under the guidance of Prof. Sofana Reka and Dr. Kumar Rajamani. During this time, I interned at École Polytechnique de Montréal with Prof. Hervé Lombaert on cortical surface analysis using Graph Neural Networks, and at Mayo Clinic under Dr. Srinivasan Rajagopalan on tracheal analysis using CT Scans for COVID-19 severity classification.

Research Interests: 2D/3D Computer Vision, 3D Reconstruction, NeRF

Email 1  /  Email 2  /  GitHub  /  Resume  /  Google Scholar  /  LinkedIn

Computer Vision

Aug '25 - Present

Lead MLE

July '22 - July '25

B.Tech in ECM

July '18 - June '22

Research Intern

Aug '21 - Dec '21

Research Intern

Nov '20 - Apr '21

Wiggle and Go! System Identification for Zero-Shot Dynamic Rope Manipulation

Robust Deep learning Model for Detection of Tomato Bacterial Spot on Novel Dataset

Patent - Hyperverge Inc

1. Method And System For Determining Liveness of a Subject. (18/891,393)
2. Method And System of Image Processing For Determining Liveness of a Subject. (18/891,650)
3. Method And System of Video Processing For Determining Liveness of a Subject. (18/891,866)

Shape analysis of complex brain surfaces

Developed an efficient Python pipeline for spectral alignment of brain mesh surfaces using reference mesh. Explored Gaussian Kernels with Graph Convolutional Neural Neworks (GCNNs) for brain surface analysis which is used for various downstream tasks.

COVID-19 Severity classification using Tracheal density

In this work, we wish to classify severity of Covid infection using Chest CT scans. Extracted tracheal region using airway extractor module in Slicer 3D software and PyRadiomics to extract 140 tracheal density features. Applied feature engineering techniques (Lasso, Boruta) and compared performance across ML & DL models to classify tracheal infection.

Federated Learning for Liver Tumor Segmentation

Created a federated learning simulation with 2 clients using flower platform where client-1 has liver and tumor data and client-2 has only liver scans. Achieved 0.793 dice score for client-2 on tumor segmentation task with a U-Net architecture model which was trained on only liver data.

Diagnosis of Portal Hypertension using semantic-segmentation

Worked on semantic segmentation of hepatic vessels using 2D/3D U-Net architectures. Implemented Domain Adaptation by first training a model on vein segmentation and then fine-tuning it for portal vein segmentation.

Multiview pose synthesis from 2D objects

Developed an end-end approach to edit pose of an object in a 2D image given the azimuth and polar values. The pipeline includes text-guided segmentation, image to 3D view generation followed by SD Inpainting and replacing the generated object in the masked image.

Chest X-Ray Pneumonia Detection

Built a Pneumonia detection model using Convolutional Neural Networks (CNN) and Transfer Learning. The model was trained on the Chest X-Ray dataset from Kaggle and achieved an accuracy of 98%.

credits to       original template