Kumar Selvakumaran Email & Phone Number

Georgia, United States

• Augmented and fine tuned versions of BERT for entity recognition and span detection from radiology reports.• Built an ensemble of the entity recognition model and R-BERT to infer Knowledge graphs from free text radiology reports.• Implemented a semantic query pipeline using vector search and community detection to extract relevant subgraphs and classification triggers which improved retrieval F1 score by 0.05 compared to LLAMA 3 70b based retrieval.

Singapore

- Developed and containerized apps based on open-source computer vision repositories for AI microservices.- Implemented scalable video understanding pipelines for action recognition with weighted logit smoothening.- Implemented model-driven relabelling and denoising for computer vision datasets and fine-tuned multi-modal models for various requirements using PyTorch.- Developed prompt-based Zero-Shot Segmentation APIs .- Adapted the joint-training strategy from the YOLO 9000 paper and conducted distributed fine-tuning of YOLOV5’s classification head on an in-house dataset using a Google Cloud Virtual Machine consisting of 4 NVIDIA Tesla T4 GPUs.

Chennai, Tamil Nadu, India

-Played a key role in developing computer vision solutions for retail analytics hosted on GCP.- Practiced the whole ML life cycle from data wrangling (image stitching using OpenCV C++ to capture a whole scene) and data annotation using Amazon SageMaker and Mechanical turk to model monitoring on GCP- Model and workflow maintenance- Developed a custom object detection pipeline to handle dense object detection in high resolution images (8k - 32k).- Provided valuable insights on building increasingly scalable solutions by understanding and observing state-of-the-art methods.

Chennai, Tamil Nadu, India

Developed a real-time face mask detection script using Keras, Tensorflow, OpenCV.

Chennai, Tamil Nadu, India

- Gained practical experience in the various aspects of social media marketing with a focus on Linkedin marketing. - Got a first-hand glimpse of how start-ups and small-scale businesses function and the various strategies used to promote their growth and outreach. - Helped in research and content creation for the official website.

Master'S Degree, Artificial Intelligence, 3.77

Coursework: Algorithms, Fundamentals of AI, Pattern Recognition and Computer Vision, Unsupervised Machine Learning, Machine Learning..

Bachelor Of Technology - Btech, Computer Science With Honors In Tensorflow And Minors In Business Analytics And Problem Solving, 3.45/4

High School Diploma, Computer Science