comicencyclo,Sandip Bhattacharjee,github

Hi there 👋

I am Sandip. I am leading Data Science and ML Engineering at Walmart Marketplace . I am also a part-time Kaggler and currently a 4X Kaggle Expert. You can find me on Kaggle here: https://www.kaggle.com/sandy1112

My Core Competencies:

Large Scale Recommender Systems
User based Personalization
Image Classification
Object Detection, Segmentation
Sentiment prediction, Span prediction, Sentiment classification
Demand Forecasting
Market Mix Modeling
Multi Touch Attribution
Test Control Modeling
Pricing & Promo Models
Store Clustering
Trip Segmentation

Techniques that I have implemented in a hands on manner:

Latent Class Models
Timeseries Forecasting
GLM, HLM
State Space Models
Decision Tree
Random Forest
Gradient Boosted & Extreme Gradient Boosted Trees
Deep Learning
- Convolutional Neural Networks
- Recurrent Neural Networks (LSTM, GRUs)
- Transformers (BERT, RoBERTa)

My 15+ years of industry experience spans accross:

Retail
CPG
Financial Services
QSR and F&B
Oil & Gas

Creating 1:1 personalization solutions is an special area of interest to me.

🔭 I’m currently working on Applications of Large Scale ML on Online Marketplace.
👯 I’m looking to collaborate on novel ways to create recommender systems using Deep Learning. Also, I like Computer Vision problems as well and always looking to collaborate on Computer Vision problems.
⚡ Fun fact: As my GitHub name suggests, I am an avid comics reader. This includes comics of all types - English, Hindi, Bengali. Also, creating my own comics strips at leisure is something I love to do.
📫 How to reach me: https://in.linkedin.com/in/sandip1006

Sandip Bhattacharjee's Projects

create_and_train_resnet50_from_scratch

The purpose of this repository is to illustrate how to create and train a ResNet50 from scratch. The actual notebook can be found on my Kaggle profile with the sample data that was used for this.

eda_and_modeling_for_largescale_imageclassification

The purpose of this repository is to show how to do EDA and modeling for a large scale image classification problem

gameofdl-withfastai

This repository provides my solution to Analytics Vidhya hosted hackathon "Game of Deep Learning". This was based purely of fastai with minimal testing.

keras_testtimeaugmentation

The purpose of this repository is to provide Test Time Augmentation abilities to Keras models for application in Image Classification tasks.

ltfs-datascience-finhack-an-online-hackathon-solution

This repository presents my solution to the recently concluded "LTFS Data Science FinHack ( ML Hackathon)" on analyticsvidhya.com . This will illustrate a simple case of averaging multiple model types to achieve higher LB scores than any single model. The private LB score achieved by this solution is 0.66085+ , which got me Pvt. LB rank of 170.

ml_interpretation_ensembling_for_strctued_data

The purpose of this repositiry is to illustrate some methods of interpreting the Machine Learning models (many times also called as Black Box models) for their so called 'Lack of Interpretability. I will also show a few methods of ensemling various models to achieve better results than any single model.

transferlearning_discriminativefinetuning

The purpose of this repository is to illustrate the implementation of Discriminative Fine Tuning for Transfer learning models in Keras.

what_a_tree_based_model_sees

This repository will provide an example of what a tree based classification/Regression model sees. This will provide an intuition behind feature engineering for tree based models. The data used in this example can be downloaded from https://www.kaggle.com/c/santander-customer-transaction-prediction/data

Recommend Projects