Write 2-3 short paragraphs about who you are and what you’re famous for ✌💁

This Card can be used for lots of other purposes, like showcasing a blog post or a new video.

This podcast is on futuristic and latest technologies that include but are not limited to Singularity, Metaverse, Blockchain, Data Science, Artificial Intelligence, Machine Learning, Immersive [Extended/Mixed/Augmented/Virtual] Reality, Internet of Things, Industry 4.0, Design Thinking, Quantum Computing, Robotics and Apps. Currently in its Season 1, the next season will feature interviews with CEOs of Tech companies.

My Top Projects

1. Applied an advanced machine learning technique, Deep Boltzmann machines as generative models, to find complex interaction patterns in medical data where it is hard to pool individual data due to data security and data protection concerns, and hence a synthetic data approach was used.

2. Demonstration of top data science tools, libraries and techniques from Uber, Netflix, Google, Lyft etc. for use in data science projects - Ludwig, GPipe, PyText, Zero Shot Learning, MLflow, Neptune, Vaex, Helix, AutoML, Michelangelo, Polynote, Manifold, Horovod, Flyte.

3. Developed a Review based Transformer Model for personalized product search to help customers base their decision on personal preferences besides product relevance.

4. Developed an Encoder Decoder based deep convolutional neural network architecture with Multi-scale-aware modules for crowd counting and generating high quality density maps.

5. Demonstrated the use of Big Data Analytics in Industrial Internet of Things with approaches such as Self Organizing Map algorithms for mass product customization towards lean manufacturing, Industrial Time Series Modeling, Intelligent shop floor monitoring, Industrial Micro grids, Monitoring machine health and intelligent predictive and preventive maintenance.

6. Used an automatic framework to minimize memory footprint on a deep learning network for video recognition. It helped reduce the overall memory requirement by 3× for the PyTorch model, with a 10% overhead in computation.