itneshkumar Goto Github PK

[email protected] linkedin.com/in/itneshkumar +91 8010112762 NCR Delhi, India

With 5 years of experience in the AI Industry problem-solving expertise in the banking and mapping industry. Utilizing NLP, LLM, Computer Vision, and Machine Learning to solve various challenges related to Text Processing and Sentiment Analysis Generative tasks, Road Object Detection and satellite Image segmentation. Specifically, I have focused on text processing, sentiment analysis generative tasks, road object detection, and satellite image segmentation, Statistical model training.

Specialize in implementing a clean architecture for deploying models in production, prioritizing efficiency and effectiveness in the process.

VKYC: This service has mainly five features with multiple modules.

• Document Data Extractor & Verification
  

  • Document Matching

    Documentation should be uploaded correctly not any wage document uploaded by the user.
    

  • Aadhaar Card
    

    • LayoutLMV2
    • Visual QNA
      Tesseract

  • Pan Card Data
    

    • LayoutLMV2
      Tesseract

  • Video Verification
    

    • Head Movement Detector
    • Eye Blinking Finger Counter

  • Signature Match Verification
    

  MinIODB

AI Service & AI Prediction : Some of the use cases of service

• Fraud Detection: Analyzing data to detect fraudulent activities.
• Up-Selling and Cross-Selling: Identifying potential additional product or service recommendations based on customer behavior and purchase history.
• Loan Interest Customer Prediction: Predicting the likelihood of customer interest in a loan using historical data and application details.
  Used Techstach:

MinIODB Model cache Model cache

ChatBot Next :

• This chatbot has been primarily trained using custom data from a banking industry survey, and its underlying framework is based on Rasa.
• And profonaity filtration word as well.

MinIODB Model cache

OCULUS Service:

• This tool streamlines media management by providing a single window for accessing various social media platforms such as Facebook, LinkedIn, Twitter, and Instagram.
• Additionally, it provides features like post sentiment analysis, summarization, language translations, and suggestions for business-related posts.
• Moreover, it facilitates campaign scheduling and posting across social media channels.

TSDR: Traffic sign Detection and Recognition

• Detecting and Recognizing up to 100 different Driver-Alerts in internal survey street image data.
• Effort reduction of data team up to 80%.
• GPU handling is used so that our model can take advantage of production resources.
• It reduces driver mistakes and suggests the all-driver alert for smoother rides. Due to being connected with GPS, it also provides the location of driver alert as well.

Privacy Keeper: Face and License Plate Detection and Blurring

• The license plate detection utilized the YOLO model, while the MTCNN model utilized the face detection.
• Specialized Indian datasets were used for training.
• GPU multiprocessor functionality was used with Tkinter for GUI.

Brand Billboard Detection and Recognition:

• The system identifies brand billboards in images or real-time webcam feeds. The recognized brand names are acknowledged by the company, allowing for training based on this information.
• The system is deployed for batch processing purposes.

Zebra Crossing Detection and Recognition:
The system detects all pedestrian crossings on the road in real time and promptly alerts the driver.

Address Parser:

• Based on entity recognition, the system identifies and categorizes words related to geographical addresses, such as village, city, and country.
• This information is useful for location searches.

• MACHINE LEARNING, NLP: AutomL pycaret (training), Langchain, Layoutlmv2, Tacotron, Hugging Face, Transformer, Rasa ChatBot

• COMPUTER VISION (Convolutional Neural Network): Training, Transfer learning with pre-trained models on customized datasets like (ResNet, VGG, Yolo, UNet, Mask-RCNN, etc.), Object Detection and Recognition, Image Segmentation and classification, Modeling with Keras, Tensorflow, and Pytorch, Image Processing with OpenCV, Matplotlib

• DATA PREPROCESSING AND VISUALIZATION: Pandas, Numpy, Pyplot, Plotly, Beautifulsoup, Boxplot, Histogram, etc.

• DATABASE: Postgres, Mysql and Mongodb, minIO, Milvus DB

• DEV TECHNOLOGY: AWS, Conda, Cuda (GPU), Docker, Kafka, PySpark, Kubernetes (Testing: unit test, pytest)

• PROJECT MANAGEMENT TOOL: GitHub and Jira

• API FRAMEWORK: Flask and Fast API, Sanic API, tkinter, Streamlit, gRPC