ThirdEye delivers end-to-end Data and AI solutions for enterprises worldwide by leveraging state-of-the-art Data & AI technologies. ThirdEye Data has over a decade of experience in transforming enterprises like Walmart, Microsoft, Southern California Edison, Amgen, British Petroleum, and a host of large and small companies worldwide.  ThirdEye has offices in the Silicon Valley – California, USA, and Kolkata, India with new offices coming up in Montreal, Canada, and Bangalore and Hubli, Karnataka.

• Applications of Artificial Intelligence, Machine Learning, and Deep Learning technologies for the real-world. One such real-world problem that is covered in this workshop is analyzing the image quality issues in the manufacturing industry.
  Using Computer Vision, we have built the below three features:
  1. Object Detection: This is an image object detection model which identifies the objects and their location in the image.
  2. Occlusion Detection: This model can detect if an object is clearly visible or occluded by any other objects.
  3. Blurriness Detection: This model tells us whether the given image is clear enough to see all the components on the pole or not.
• Understand the complete end-to-end AI Processes:
  • Model Development
    • Define business and technical requirements 
    • Identify various algorithms as per the definitions
    • Perform a lot of experimentation and iteration with all available data sets 
    • Explore data to understand the underlying correlations between different variables
  • Model Training
    • Information gathered from the data exploration process will help us to choose the best model
    • Develop the right model by iterating through its  parameters and tuning it to get optimal outputs
    • Split the available data into three parts – training, validation, and testing
  • Model Deployment
    • Validate the models based on both human feedback and outcomes analysis
    • Deploy the models for downstream consumption
    • Keep checking the model outcomes for model percisicon, accuracy and drift purposes. 
    • Retrain as and when needed
  • Data Engineering aspects of Model Development
    • Data Collection: Various Data sources (mobile apps, websites, web apps, microservices, IoT devices, etc.) are instrumented to collect relevant data.
    • Data Ingestion: All this data gets collected into a Data Lake.
    • Data Preparation: It is the extract, transform, load (ETL) operation to cleanse, conform, shape, transform, and catalog the data blobs and streams in the data lake; making the data ready-to-consume for ML and store it in a Data Warehouse.
  • Model Validation
    • K-fold cross-validation with an independent test data set
    • Leave-one-out cross-validation with an independent test data set
    • Perform a train/validate/test split on the data
  • Feedback Loop Analysis
    • This is the process of leveraging the output of an AI system and corresponding end-user actions in order to retrain and improve models over time.

Take-aways 

• Understand what it takes to develop AI models for the real-world
• Ability to comprehend the AI processes end-to-end

Pre-Requisites

• Knowledge and experience in 
  • Python, 
  • Jupyter Notebooks
  • Statistics
  • Analytical Skills
• A deeply curious mind!
• Adequate domain knowledge