Prasangi Bhaskar Tarun

Prasangi Bhaskar Tarun

Data Scientist

Followers of Prasangi Bhaskar Tarun419 followers
location of Prasangi Bhaskar TarunVishakhapatnam, Andhra Pradesh, India

Connect with Prasangi Bhaskar Tarun to Send Message

Connect

Connect with Prasangi Bhaskar Tarun to Send Message

Connect
  • Timeline

  • About me

    AI Engineer@IBM ( MTech at IIT Ropar)

  • Education

    • Indian Institute of Technology, Ropar

      2019 - 2021
      Master of Technology - MTech Electrical and Electronics Engineering
  • Experience

    • Pivotchain Solutions

      Aug 2021 - Sept 2023
      Data Scientist

      Intelligent Vision for your Intelligent Automation of safety, security, and process needs. Pivotchain Solutions is an advanced Artificial Intelligence company that delivers cutting-edge, real-time video content intelligence globally. With our event-based and hardware-agnostic design, the company’s video analytics software tools, and solutions are available under the RAVEN AI brand across major industries and verticals, including Logistics, Homeland Security, Aviation, or Critical Assets.• Integrated Grounded DINO with an annotation tool, enhancing object detection efficiency and accuracy while bridging the gap between human understanding and machine perception.• Developed an automatic annotation system by integrating SAM (Segment Anything Model) with the annotation tool, generating segmented data from bounding box data.• Implemented an Active Learning Model for Alert Classification, optimizing the alert system and improving classification accuracy.• Created a CNN + LSTM-based Human Action Recognition in Videos (Video Classifier) for accurate recognition of human actions in video sequences.• Enhanced Classifier models' performance by 15% using Cosine Convolution techniques in PyTorch and TensorFlow.• Built a PyTorch-based Adversarial Object Detection model to address class imbalance issues during training by generating synthetic data.• Developed an automated data generation process from videos using FFMPEG and an SSIM loss function, significantly reducing processing time.• Improved the performance of the Face Recognition pipeline by 20% and reduced false detections by 50% through the integration of Face alignment and Side face detection techniques. • Developed an Office Monitoring System using Docker, PyTorch, and Flask API to detect and alert unauthorized entries, enhancing security.• Standardized image data file names using EasyOCR for improved organization and consistency. Show less

    • IBM

      Sept 2023 - now
      AI Engineer
  • Licenses & Certifications