Prateek Kumar Verma
Computer Science and Engineering
Indian Institute of Technology, Delhi
Academic Details
Year Degree Institute CGPA/Percentage
2013-2017 B.Tech in Computer Science Indian Institute of Technology 9.09/10
(Expected) and Engineering Delhi
2013 Class XII, CBSE Nalanda Academy, Kota 91.2%
2011 Class X, ICSE The Cathedral and John Connon School, Mumbai 93.8%
Scholastic Achievements
All India Rank 68 in Indian Institute Of Technology Joint Entrance Examination (JEE Advanced-2013).
Received the Summer Undergraduate Research Award (SURA), awarded by Industrial R&D Unit, IIT Delhi
for exceptional research in the field of Computer Vision.
Awarded merit scholarship and certificate for being in Top 7 % of the institute in semester I
Internships and Major Projects
Estimating Size of Apache Lucene Indices LinkedIn Corp, Bangalore, India
Summer Internship May - July, 2016
Worked on EzSearch Tool, an internal tool at LinkedIn Corp. which automates the process of creating search services
over a given data set.
Analysed the working of Apache Lucene which is a high-performance, full-featured text search engine library.
Designed a model to approximate the structure of inverted and forward indices of Lucene with high accuracy.
Implemented a robust method to estimate the size of structure for any given data set.
Used Apache Pig over the Apache Hadoop framework to make the method scalable to large data sets.
Integrated the functionality into the existing pipeline and performed exhaustive tests.
Behavior Based Spam Filtering Models Prof. Huzur Saran
IIT Delhi Bachelor’s Thesis July, 2016 - Present
Designed and implemented three novel user profile models to encode the behavior of an email account for detection
of unusual activity and compromised account.
Postfix logs were parsed to extract relevant information like recipient lists and time stamps.
User social cliques were formed using parsed logs of last several months and compared against new email activity.
Recipient Frequency graphs were maintained and compared against recent activity using Hellinger Distance metric.
Cumulative Mail Frequency were logged over a week’s window to detect any out-burst of emails.
All of the models were combined to increase detection rate and reduce false positives.
3D Reconstruction on Mobile Device Prof. Subhashis Banerjee
Summer Undergraduate Research Project January - November, 2015
Developed a mobile app for near real time 3D reconstruction of monuments/objects.
Used accelerometer, gyroscope, magnetometer (IMU sensors) data for rotation and translation matrix estimation.
The estimates are improved by using local regression and visual tracking.
Used a Kalman filter, dense and sparse optical flow to improve the extrinsic camera parameters.
Designed a 2-point algorithm to reduce computational complexity.
The challenge was to complete dense 3D reconstruction in near real time on mobile devices.
Other Projects
OCR Character Recognition Using Graphical Models Prof. Parag Singla, October, 2016
Designed an undirected graphical model to recognize individual character of the words from images. Implemented
two-way sum-product message passing algorithm over the clique tree for inference. Also, implemented Gibbs Sampling
for the graphical model.
Object Removal through Context Aware Image Resizing Prof. Prem Kalra, September, 2016
Implemented Seam Carving method of re-sizing an image. Computed the seams to remove by finding the path
of minimum energy cost from one end of the image to another using dynamic programming. Used general Hough
transform to identify a particular object for removal.
Network Based Multi-player Game with AI Prof. Huzur Saran, March - April, 2015
Developed a network based multi-player shooting game using OpenGL library for graphics and UDP protocol. The
game is P2P with Artificial Intelligent players, thus allowing endless continuity of the game during network outages
of one or many of the players.
Digit Recognition using Neural Network Prof. Parag Singla, February, 2015
Trained a Neural Network with one hidden layer comprising of 100 units to identify digits in an image. Used
stochastic gradient version of the back propagation algorithm with a variable learning rate.
Operating System Virtualization Prof. Kolin Paul, April - May, 2016
Implemented virtualization in JOS, a Unix-like operating system designed in an exo-kernel style. First, essential
components like system calls, memory management and user environment were implemented. Then, JOS was run
as a guest operating system on JOS by running privilege instructions through trap-and-emulate virtualization.
Cloud Storage System Prof. Huzur Saran, February 2015
Implemented a cloud storage system allowing users to sync files with the server and share files with each other.
OpenSSL was used to ensure encrypted file transfer. Used FTP and TCP-IP for sync and transfer. Implemented
data de-duplication to minimize server disk usage.
Relevant Courses
Computer Science:
Machine Learning, Probabilistic Graphical Models, Digital Image Analysis, Artificial Intelligence, Operating Sys-
tems, Parallel Programming, Computer Networks, Programming Languages, Computer Architecture, Analysis &
Design of Algorithms, Data Structures & Algorithms, Discrete Mathematical Structures, Digital Logic Design
Mathematics and Electrical Engineering:
Calculus, Linear Algebra, Intro to Electrical Engineering, Probability & Stochastic Processes, Signals & Systems
Technical Skills
Programming Languages: C, C#, C++, Java, Python, SML, Lex, Yacc, Prolog, MySQL, HTML, PHP
Extra Curricular Activities
Runner up in Code.Fun.Do - 2015 organized by Microsoft amongst students from different colleges from India.
Pursued an internship at Khushiya Sewa Sanstha NGO which works towards the upliftment of the blind and the
underprivileged in Mumbai.
Member of the Student Mentorship Program which ensures smooth transition of new students into college life.