Momin Patel

Proficient in harnessing the power of computer vision, including cutting-edge techniques like the YOLO (You Only Look Once) model, to extract actionable insights from complex datasets. Experienced in statistics and adept at exploring intricate visual data, integrate computer vision models like YOLO for object detection and recognition tasks. Acquiring expertise in computer vision and its applications, with a focus on enhancing AI models like GPT-3.5 for visual tasks. Effective communication with AI models, including GPT-3.5, to drive innovative projects and data-driven decision-making in computer vision and AI. Accomplished Business Analyst and Client Service Associate transitioning into Artificial Intelligence and Data Science. Experienced in managing global teams and collaborating with C-Suite Executives to deliver impactful computer vision solutions.

Data Science & Artificial Intelligence. In Collaboration with UTM - Malaysia 360 DigiTMG, Hyderabad

Project: Automated Egg Detection for Poultry Industry

Objective: Create an efficient, error-reducing, and cost-effective automated system for the poultry industry.

Data Collection and Preparation:

• Collected 100,000 diverse images of white and brown eggs, including non-egg objects, for robust training.
• Utilized Roboflow for data preprocessing and augmentation to ensure data quality.

Model Development and Enhancement:

• Developed the YOLOv8 model with a medium configuration, achieving an initial mAP of 83% for egg detection.
• Expanded the dataset to 150,000 images for improved accuracy.
• Upgraded to YOLOv8 Large, achieving an impressive 90% mAP for precise egg detection.

Success Criteria:

• Business Success: Reduced egg counting time by over 30%, enhancing efficiency.
• Machine Learning Success: Achieved ≥90% accuracy in egg detection.
• Economic Success: Saved >$1 million by improving efficiency and reducing counting errors.

Cloud-Based Training and Deployment:

• Leveraged AWS for scalable and efficient model training.
• Deployed the system in real-time using Apache Kafka, enabling instant and accurate egg identification in poultry operations.

Project: Healthcare Chatbot Development with ChatGPT & LLM

• Developed an intelligent healthcare chatbot using ChatGPT and Large Language Models (LLMs) to provide user-friendly healthcare support.
• Conducted thorough testing with multiple LLMs, including Palm, Palm2, MedPalm2, Llama2, Langchain models, and ChatGPT, to assess their performance in healthcare tasks.
• Fine-tuned using Palm2 LLM model to achieve an impressive 80% accuracy rate for precise responses to health-related queries.
• Deployed the chatbot on the AWS cloud platform to ensure scalability and widespread accessibility for healthcare inquiries.
• Configured the chatbot to perform specific downstream healthcare tasks, including symptom assessment, medical advice, appointment scheduling, and prescription inquiries.
• Designed a user-friendly interface to facilitate easy interaction with the chatbot, empowering individuals to receive personalized and accurate healthcare guidance.

Project: Fraud Detection in Procurement on Big Data

• Highlighted the substantial global loss of $3.5 trillion annually due to procurement fraud, representing 4-8% of an organization's procurement expenses
• Focused on developing an ML model with the primary goal of enhancing fraud detection in procurement processes while minimizing organizational costs.
• Conducted automated exploratory data analysis (EDA) using Pandas Profiling and SweetViz.
• Employed AutoML techniques with TPOT, identifying gradient boosting as the best-performing model with an 85% accuracy rate.
• Achieved an accuracy improvement to 91% through hyperparameter tuning of the Random Forest model.
• Demonstrated the substantial potential for cost savings by applying the developed model to just 10% of total procurement transactions, estimating global savings of about $350 billion annually.
• Successfully deployed the fraud detection model using Flask and Google Cloud Platform (GCP), enabling real-time application in procurement processes.