I am a Graduate Student in Computer Science at Binghamton University. I have 2.5 years of experience in enterprise software application developement. I am actively seeking full time opportunities starting Jan 2019.
I am a Graduate Student in Computer Science at Binghamton University. I have 2.5 years of experience in enterprise software application developement. I am actively seeking full time opportunities starting Jan 2019.
Developed recommender system for partially completed matrix using Single value Decomposition. Entire dataset was given with missing user, item, rating format(Similar to MovieLens dataset). Applied matrix factorisation technique and reconstructed original matrix from latent features. RMSE/MAE reported was 0.5(approx) for all the predictions made. Used surprise and Python 3.
Modified Xv6 kernel to support MLFQ scheduling through 2 queues which had promotion/demotion logic to switch a process between queues. Thread API provided support for thread creation through 'Clone', 'Join' system calls so that multiple threads can share same memory. A new ‘xvsh’ shell was developed to run processes in background and handle zombie processes. All these functionalities were developed by providing multiple system calls. C and QEMU were primarily used.
Monolithic 3 tier MVC based application was developed to support student registration in an university. Mutiple stored procedures, functions, triggers were implemented as part of db implementation. Data was fetched into model through JDBC and RefCursors. Servlets handled controller features to link view and model. Dynamic View was based on JSP, JSTL, Javascript and static UI related technologies (HTML, CSS, Bootstrap).
Carried out decoupled multi stage simulation for an out-of-order APEX pipeline with dedicated Integer, Division and Multiplication Units. Each instruction had to go through FETCH, DECODE, EXECUTE(Div, Int, Mul), MEMORY, WRITEBACK stages. Bubbles were generated as part of instruction stalling. Reorder buffer and Load-Store queue were implemented to further reduce execution cycles.
Designed and implemented multinomial Naïve Bayes classifier with Laplace add-one smoothing on Metsis et al. paper dataset to predict spam/ham e-mails. Stop words removal were done through NLTK to further improve accuracy. Smoothing prevented probabilities getting evaluated to 0 which in turn boosted accuracy. Entire calculation was log-based to prevent underflow error in any case.
Developed MCAP LR algorithm with L2 regularization on Metsis et al. paper dataset. Used gradient descent algorithm for convergence. Compared two model predictions by implementing Perceptron training rule based classifier. Perceptron training rule was induced for Perceptron classifier whereas Sigmoid function was used to model dataset for MCAP LR to test for convergence.
Developed micro services for Network Insights Advisor DeepScope for Cisco's DCNM. Containerize service through Docker and allow container orchestration through Kubernetes. Kafka cluster was configured to facilitate messaging among multiple services. Synchronous calls were thrift/rest based to enable UI communication. Front End developement mainly included chart and graph reports in D3.js.
Designed Forms/Dialogs/Tables using HTML, CSS, JS libraries and Bootstrap as part of UI implementation. REST based API architecture involved creating new endpoints for client-server communication. Created Resource/Controller/Handlers in Java to consume rest based endpoints. Refactored legacy code and reduced N-path complexity to enhance code performance and maintainability.
Designed and implemented stories in each Sprint cycle. Refactored existing code. Handled production issues and worked with L1/L2 support team to provide hotfixes. Integrated Jasmine/Karma modules into existing UI code base.
Designed ERP(Admin module) and implemented code from scratch. Completed unit, functional and integration testing for the entire module. Front end development involved working with HTML, CSS, JS, JQUERY & server side code was implemented using Java7. Database used was MySQL.