1. ABHISHEK KUMAR
a35@buffalo.edu
(716)525-9906
www.linkedin.com/in/ubkumarabhishek | https://github.com/abhishek-ub
EDUCATION
MS Computer Science, University at Buffalo, The State University of New York GPA: 3.75 June 2016
Bachelor of Technology, Indian School of Mines, Dhanbad, India (Graduated First Class) May2013
TECHNICAL SKILLS
Programming Languages Database Web Technologies Tools Web Server
C C++ JAVA Python
MATLAB programming
MySQL PL/SQL
Oracle 11g
JavaScript JQuery HTML
CSS XML
Juint , Apache Nutch
Apache Solr ,Scrapy
WebLogic , Tomcat
PROFESSIONAL EXPERIENCE
ORACLE Financial Services Software, Bangalore (India) Associate Consultant Sep 2013-July 2014
Served as key contributor in Product Customization for Bank Misr Lebanon. Developed module for product
“FLEXCUBE” handling Retail Banking and tracking overdraft of customer in a branch.
Implementation Consulting, involving software design strategy to incorporate Bank’s core banking needs into the product
and came up with Functional and Technical Design Documents.
Used Oracle Business Intelligence (OBI) tool to generate graphical analysis and report for consumer lending and other
modules of the product.
Using PLSQL, Java script and XML developed Backend (database) and frontend Application to be deployed in production.
ACADEMIC PROJECTS
Distance Vector Routing Protocol (Dec 2014)
Implemented Distance Vector Routing Protocol on top of servers (which behaved as routers) using Bellman Ford
Algorithm. The protocol behaved as simplified version of RP used in Unix systems to build routing table at each router.
Reliable Transport Protocols (Nov 2014)
Implemented three reliable data transport protocols - Alternating-Bit, Go-Back-N and Selective-Repeat in a given
simulator. Analysed and compared their performance.
File Sharing System (Oct 2014)
Develop application(C socket programming) to communicate over network and share file among connected peers using
TCP connection. The application supports commands like upload, Download to request file from server.
News Personalization Engine www.prefreader-prefreader.rhcloud.com (Nov 2014)
Developed a news search engine which provides content to users based on their preferences and reading behavior.
Leveraged Apache SOLR and its features to index multiple news corpus to provide query suggestions.
Implemented Collaborative Filtering to match a user with similar users and suggest articles read by these similar users.
Application suggests Articles about trending topics in area of user's interest also supports Spell suggestion, highlighting.
Deployed on PAAS -"Open shift"- An Open Hybrid Cloud Application Platform by Red Hat.
Information Retrieval System (Oct 2014)
Developed an information retrieval system (java) which parses a Reuters corpus of newspaper article (RCV1), Tokenize
and generates Inverted indexes, which is used for Boolean Queries retrieval.
Implemented vector space model (tf-Idf) and OKAPI BM-25 model for ranked retrieval of documents on user query.
Computer Vision and Image Processing (Fall 2014)
Applied photometric stereo(Matlab) to construct 3D model by calculating surface normal and height map of images
Segmented object from image using K-means clustering and projected segmented image on different background.
Semantic labelling of image segments using SVM classifier (Libsvm MATLAB library) (Dec 2014)
Labelled segments of images in “Stanford background dataset” with semantic classes of “sky”,”water,”grass” etc.
Used LibSvm’s Matlab library to train the SVM classifier, implemented 5-fold cross validation and grid search to get
best average accuracy greater than 50% and in class accuracy greater than 85% for “sky” and ”grass” class.
MPPT Algorithm for PV cells (Dec 2012)
Implemented and compared response of various maximum power point tracking (MPPT) algorithms for PV modules.
Used data from Indian meteorological department to train neural networks for tuning controller.
Independent Projects
Developed Web app at UB-hackathon (Nov-2014) which provides the user, sector wise list of companies expected to offer
most rate of return and expected return on a dollar invested.
Implemented Markov Random chain (MRC) model over 4 years of opening and closing stock prices data for different
companies collected using Bloomberg API to predict the expected return.