SAHASRA KEERTHI

AI-Powered Video Generation from Text: Advanced Media Production for RTV

• Developed an AI-powered text-to-video system for RTV that converts Telugu text inputs into fully synchronized videos, complete with realistic animations and lip-syncing, enhancing the storytelling experience and viewer engagement.
• Leveraged advanced AI technologies to automate and optimize video production processes, significantly reducing production time and costs while maintaining high content quality.
• Upheld stringent ethical standards in AI usage, achieving a high viewer satisfaction rate through transparent and responsible technology practices, thereby contributing to advancements in media production technology.

AI-Powered Small and Medium Business Loan Prediction: Advanced Financial Assistance Tool

• Developed the "Small and Medium Business Loan Prediction" system using LLAMA2 on AWS Bedrock, which predicts loan eligibility and amounts for SMBs based on key financial and personal data including Aadhaar and PAN numbers, CIBIL scores, and bank statements.
• Implemented an automated application form feature that pre-populates fields with user-provided data, allowing users to download, review, and complete the remaining sections, thereby streamlining the loan application process and improving user convenience.
• Ensured rigorous adherence to data privacy and security standards, safeguarding sensitive information through advanced security measures, thereby enhancing trust and reliability among users while supporting ethical AI practices in financial services.

Object Detection POC Development Using YOLO-NAS for Textile Industry Quality Control

• Engineered a Proof of Concept (POC) leveraging YOLO-NAS, an advanced neural architecture search algorithm, specifically tailored for object detection tasks.
• Adapted the model to accurately identify various defects and use cases within the textile sector, enhancing quality assurance processes.
• Conducted rigorous training and validation of the model on a custom dataset, achieving a marked improvement in precision and recall metrics.

Attendance management system for Access control management

• Developed a comprehensive attendance management system leveraging FastAPI and Streamlit for efficient employee management and attendance tracking.
• Implemented facial recognition features using OpenCV, dlib, and MTCNN for real-time employee identification and attendance recording via webcam and high-resolution Hikvision cameras.
• Designed and deployed Fast APIs for functionalities including user login, employee registration, fetching employee details, and retrieving attendance records.
• Integrated MongoDB for robust data storage of employee information and attendance logs, ensuring secure and efficient data retrieval and management.
• Utilized Docker for containerization and Kubernetes for orchestration, facilitating a seamless development and production environment setup.
• Leveraged Raspberry Pi and Coral Accelerator for edge computing capabilities, enhancing the system's performance and real-time processing efficiency.
• Deployed the system on cloud infrastructure to ensure scalability, availability, and reliability of the application.

• Hands-on experience in using OpenCV for image data preparation, extensively utilized cv2, NumPy(basic array operations), Pandas, and DataFrames in performing preprocessing and exploratory data analysis throughout the learning process
• Efficient in writing and optimizing the code in Python, experience on applications using Python integrated IDEs Jupyter Notebook , Spyder and VScode
• Experience in working with Computer Vision projects, building models with MobileNet,Efficientnet classifiers using Keras, Tensorflow and Pytorch - DKYC solution (Reliance JIO)
• Worked on object detection and recognition for rack/shelf detection for a retail store and recognition of object and details trained on many labels - IVY Mobility project (Reliance Retail).
• Experience on working with EDGE devices(jetson NANO, raspberrypi) and compressing various deep learning models (classifiers, object detection, FR, OCR) by Post training Quantization and TensorRT and inferencing them on the edge devices - Model conversion (JUKSHIO)
• Worked on Intelligent Access control system using facial recognition to provide security in a given environment with CL/CI/CD by using Edge devices to run machine learning models to perform on device operations and give the response in real time. - EDGE FR (SUKSHI)
• Worked in CARTZY, which is an autonomous retail solution as Hardware engineer in microcontroller programming to retrieve data from various sensors - CARTZY (Reliance Retail)
• Worked in IKA(Intelligent Kirana Automation), which is an autonomous retail solution as Hardware engineer in retrieving data from camera and various other sensors - CARTZY (Reliance Retail)
• Worked on Watermark/Stamp removal behind the text in a document, to remove the Text Recognition accuracy.-Driving License (DKYC)

• Worked on yawn detection using DLIB facial landmarks and face detection using MTCNN - EAGLE EYE (JUKSHIO)

• Worked as hardware and software verification tester of Bibox kits.

• Learnt about sensors, ignition, cooling and transmission systems in detail and worked on basic servicing along with wheel alignment.

• Trainee of 1 month, learned basic servicing of the automobile (car), sensor systems, ignition systems, cooling system and transmission system.