Amit Chandak

  • Bengaluru , India,

Research Engineer @ Amazon


  • Contribute towards the growth of an organization which provides a learning and challenging  environment to work.
  • Learn, explore and grow in the field of Machine Learning, Computational Advertising, Big  Data

Work History

Work History
Sep 2013 - Present

Research Engineer in Machine Learning Team , Amazon

  • Worked on a project  to enable sellers by giving out loans. The goal of was to increase the seller selection in the portfolio while keeping the loss at a pre-specified minimum. Increased the existing regression model AUC from 0.75 to 0.82
  • Currently working in platforms team  implementing K-Means clustering algorithm in a scalable way
    • Designed and implemented initialization module which selects the initial k centroids.
    • Implemented KMeans++ algorithm in a scalable way
    • Benchmarked performance of our implementation with other standard ML Libraries
    • Prototyping different techniques for suggesting 'k' value for a given dataset and also suggesting better convergnce criteria
Oct 2011 - Sep 2013

Research Engineer in Analytics Team , Komli Media

  • Audience Modelling
    •  Goal was to modify bid amount in RTB according to users affinity towards the ads for which bid request is received.
    • Propsed and prototypes matrix factorization solution which improved CTR by 2%
  • Explore-Exploit algorithms for new ads created in system
  • Analyzed and improved model performance which was used for evaluating the inventory for display ads in Real Time Bidding
  • Introduced new feature "day-part" in improving model relevancy for Real Time Bidding (RTB)
Jul 2010 - Oct 2011

Research Engineer in Analytics Team , Guavus Inc

  • Worked on URL Categorisation problem
    • Aim was to classify given webpage into predefined categories using only the URL
    • Prototyped and analysed different approaches based on Naive Bayes and SVM
    • Included different features like n-gram, synonyms,etc..
    • Proposed an architechture for the system based on above approaches
  • Worked on  problems related to predicting network traffic across ISP’s 
May 2007 - May 2008

Software Engineer , Infosys

  • Completed 2 months training for new joinees with A+ grade 
  • Worked as Software Engineer in Infosys on a banking project.
  • Familiar with Infosys's banking software Finalce.


Summer Internship at IBM-India Research Laboratory , May-09 to July-09

  • Was offered an internship for 3 months during my Masters in Analytics team. My mentor was Dr. Om D Deshmukh
  • Automatic Quality Monitoring of Call Center Chats
  • Goal of the project was to build a system which would analyze the quality of textual call center chats
  • Analysed initial approach using bag-of-word features and naive bayes
  • Incorporated new features like dividing the whole chat in segments using Conditional Random Fields


2008 - 2010

M.Tech from Indian Institute of Technology, Madras

Relevant Course Work

  • Advance Algorithms
  • Mathematics For Computer Science
  • Data Mining 
  • Pattern Recognition
  • Knoweldge Representation & Reasoning

M.Tech Thesis  

Support Vector Regression Using Active Learning under Dr. C Chandra Sekhar

    • Active learning has become a popular paradigm for reducing the sample complexity of large scale learning tasks. In context of Support Vector Machines, active learning can be used to speed up chunking algorithms. The aim of this project is to implement and improve the existing active learning approaches and use the resulting system for autoassociation classification and regression tasks.
2003 - 2007

B.Tech from Dhirubhai Ambani Institute Of Infomation And Communication Technology, Gandhinagar

B.Tech Project
Named Entity Extraction using Hidden Markov Model under Dr. Abhinay Pandya

  • Built a system using HMM, which would recognize and label the instances of named classes in textual environments and its accuracy was tested on Brown Corpus. The  system was developed in Java.


  • Secured an AIR of 189 (98.95 Percentile) in GATE-2008 among 18,224 students in Computer Science stream.

Relevant Skills

  • Big Data : Familiar with Map-Reduce framework and programming in Java and Pig 
  • Scripting Languages: Familiar with Perl , Python , Pig scripts for Hadoop , Octave
  • Tools : Scikit , Mallet, Weka, R