Top Skills and features used:
Python, FastAPI, Uvicorn, Transformers, Tokenizers, Groq API, Image to Text LLMs, OCR, Computer Vision, Threading, Parallel processing, ML, DL, YOLO, Ultralytics, model training, Huggingface, Postman, Whimsical, Roboflow, google colab
Key Achievements:
- π Rampsure Project:
- π± Led the Development of the Rampsure Application: Spearheaded the design and development of the Rampsure app to streamline cattle insurance claims, enhancing accessibility, accuracy, and efficiency in the claims process for paravets.
- π€ Optimized Claim Processing with AI-Driven Features: Integrated AI algorithms to analyze feature similarities between live and dead cattle images, reducing fraudulent claims and ensuring the accuracy of insurance approvals.
- π AI Verdict Report Generation with Visual Insights: Developed an AI-powered verdict report that visualizes feature comparisons through graphs, providing a clear percentage of similarity and a confidence score to determine if the claimed cattle matches the live cattle.
- π‘ Enhanced Data Quality and Remote Accessibility: Created features like "data mode" and "claim mode" for seamless remote data entry and claim submission, enabling paravets to upload high-quality images and manage claims from any location.
- βοΈ Automated Workflow for Cost Reduction and Efficiency: Automated cattle registration and claim processing, significantly lowering operational costs, reducing manual paperwork, and increasing the paravets' capacity to handle more cases daily.
- π§ͺ API Testing for Rampsure Project: Conducted rigorous API testing for the Rampsure application, ensuring robust performance, data integrity, and seamless integration with the app's features for reliable operation.
- π Client-side project: Cattle Feature Extraction and Classification:
- π Innovated and streamlined the annotation process and polylines creation based on extensive R&D, resulting in a 25% reduction in annotation time.
- π₯ Orchestrated data preprocessing for annotation, optimizing task delegation among the team, fostering a collaborative environment that led to a 30% increase in overall productivity.
- πΈ Processed and meticulously curated a diverse dataset of 20,000 cattle images, achieving a 95% accuracy rate in live and deceased cattle annotation.
- π― Executed Image Annotation on Roboflow, demonstrating proficiency in preparing over 50% of the dataset for the seamless application of the Yolov8 algorithm.
- π Conducted regular team training sessions on advanced annotation techniques, contributing to a 20% improvement in team members' annotation accuracy and efficiency.