Supratim Haldar
Manager - Data Science at Head Digital Works Pvt. Ltd. (Cricket.com, Fanfight.com, Ace2Three.com)
Bangalore, India
- * Working as a Data Scientist at Head Digital Works Pvt. Ltd., a pioneer in online gaming with a rich portfolio of cricket.com, fanfight.com and ace2three.com. At present, leading the Data Science team of cricket.com.
- * Design and develop Criclytics, a predictive model framework to predict outcome of cricket match and performance of players.
- * Work with engineering team to build and deploy the predictive models as API.
- * Earlier, worked as a Solution Architect for Oracle Corporation, building products with advanced predictive capabilities.
- * Sound knowledge of Statistics and Machine Learning Algorithms.
- * Ranked among top 10% answerers on Python in StackOverflow.
- * Hobbies: Cycling, running, photography and reading.
- * Trivia: The profile picture on the left is generated with my implementation of neural-style-transfer algorithm, transferring style from Vincent van Gogh's famous self-portrait to my photograph 😊
Skills
Summary of my skills in the data science, machine learning and deep learning area.
In-depth understanding of key ML and DL algorithms
Python (including Numpy, Scipy, Pandas, SKLearn)
Deep Learning, inclusing Tensorflow and Keras
Visualization: Matplotlib, Seaborn etc.
Customer and Marketing Analytics
R (Including tidyverse and tidytext)
Natural Language Processing
Statistics and Probability Theory
Communications and Soft Skills
Project Portfolio
A ML prediction model to predict winner of ODI cricket (sports) matches before the match starts. Prediction accuracy outperforms that of Google.
Identify the artist from a painting, built with Convolution Neural Networks
A web-application for DeepArtist, built with Flask and deployed on Heroku
Generating art using Deep Learning
As part of an online challenge, built a classifier which can predict presence of an outdoor scene (eg: mountains, glaciers etc) with an accuracy of ~97% on train data and ~94% on test data
As part of Kaggle competition, built a classifier capable of predicting whether an image contains a columnar cactus, with AUROC of upto 0.9984 on unseen test data
An analysis of accident data on roads of Bangalore
Recommend movies to user based on taste of other similar users (developed from scratch)
A deep-dive into churn analysis in telco industry and predicting next possible churns
Recommend movies to user based on different parameters of movies (developed from scratch)
My implementation of key Machine Learning algorithms from scratch in Python
Collecting data from the world wide web