Divay Jindal
Coder / Thinker
Divay Jindal
Coder / Thinker
Hey!! This is Divay Jindal. I am Machine Learning Enthusiast and loves to play with data, plot distributions and create random variables out of it. I have pretty good experience of this field. I also have worked on building statistical models on enormous amount of data and even put them into production.
Work Experience
-
Software Developer
July 2017 - June 2018
BookMyShow- ARIMA modelling and Poissonregression for forecasting pricecategory wise ticket sale in python with 70% accuracy.
- CF over user clustering forr ecommending movies.Improved conversion metric by 4% and CTR by 7%.
- Dynamic Pricing based on demand.Incorporated dynamic regression for forecasting sale in next time slot. Overall increase in penetration by 7.9%
- Naive bayes over events metadatato identify wrongly taggede vents.
- RBM for recommendation on events using Tensorflow backend.
- Model serving API’s in php.
-
MACHINE LEARNING CONSULTANT
Oct 2016 - April 2017
THE SOCIAL STREETDeveloped an application that would detect gender and age of a customer providing intelligent retail analytics . HAAR classifiers and HOG detectors turned out to be the saviour for this reason being their optimised time and processing requirements.
-
DATA SCIENCE INTERN
May 2016 - July 2016
CUBE26A generic query generator was developed. The query was developed using Word2Vec which optimized the search time of Elasticsearch framework by 200 ms.
-
HEAD GYAANSAGAR
Nov 2013 - March 2017
NGO SILCHARFounded a non profit organization aiming at the skill development of the the children. Skills included giving basic knowledge of computers , life skill , basic mathematics etc.
Education
-
NATIONAL INSTITUTE OF TECHNOLOGY
2013-2017
SILCHARFounder Data Science club. Tought Linear Algebra and Machine Learning to the fellow students. Represented in college in National level coding championship .
-
DLF PUBLIC SCHOOL
2003-2013International Mathematics Olympiad scholar
Anything Optimizer | Using Swarm Intilligence
In parallel to Gradient Descent,FTRL,Genetic Algorithm with are used to optimise and given cost function following some heuristic, this custom written optimizer does the same thing exploiting the newtonian laws of motion. OpenSource and free to use.
Word2Vec Generic Query on Elastic
Implemented Word2vec to create a more intelligent query on elastic making use of sharding. This intelligent query framework decreased the search query time by 20ms as compared to traditional routing mechanism followed by elastic framework.
Retail Intelligence with openCV and Scikit-Learn
Used HAAR and HOG for fast human recognition over frames of video. Further tracked the movement of customer using Kalman Filter. Using HAAR over face and further face alignment usinhh 128 facial key points identified the gender on the customer in the video. Pitched this product to Samsung and Apple.
Genetic Algorithm for solving Sudoku
Using paradigm of survival of the fitted solved the Easy, Medium and Hard Sudoku problem implemented in cpp. The algorithm took avg time of 2.15sec to solve Hard problem.
Condense
My implementation for gzip compression algorithm. Still trying to optimize on the algorithm. Free to use and discover.
QUESTION CLASSIFIER USING SUPPORT VECTOR MACHINE
First step machine learning where implemented the question classifation based on pre defined categories using support vector machine over unigrams and bigrams tfIdf bag of words.