Career Profile
A zeal for programming keeps me driven to seek for software challenges and new ideas. In my various attempts to solve problems whether alone or in a team my focus remains on good design and understanding of the logic and theory behind the problem. I have knowledge Java, Python, Scala and across different frameworks and platforms and look forward to constantly growing my skills and applying them to produce wonderful products in future.
Experiences
* Implemented backend software integrations for Amdocs Ensemble framework. (Unix/Linux shell scripting, Java)
* Developed robust software modules to analyze distributed environments and their load-balancing.
* Software and scripts to track and maintain Linux environment configurations using Git, as part of Innovation Team effort. (Java)
* Developed a Java based RCP application to implement business crawler, to procure, track and update data from online product catalogues.
* Developed a mini-search utility to allow users to search for products for brands from data that was gathered.
Responsible for grading, evaluation and consultation for projects, assignments and homeworks for the Computer Science I course for Non-majors under Dr. Karen Doore.
Projects
- Developed a project for mining information on students and alumnus from data collected from social and online portals.
- Implemented RESTful services for data parsing module, search, login, authorization, advanced search, profile creation modules
- Implemented internal controller modules for authorizations, database operations and API integrations
- Developed a search engine over 120,000 documents relevant to automobiles (brands, sales, blogs, news).
- The documents were crawled and stored using Nutch 1.4
- The indexing system was implemented using SOLR 4.10 platform. Implemented PageRank with TFIDF and Okapi weighting scheme
- Query time relevance was also implemented to test the effect of HITS algorithm on the results, which clearly showed improvement
- Online shared canvas that can be edited by many users at a time, developed by leveraging Node.js and sockets.io
- Implemented Node.js project setup, Socket calls module.
- Implemented Vector Space model based relevance models using TF-IDF and Okapi(BM25) weighting schemes, and analysed the relevance results obtained.
- Extensive use of TreeMap data structure to store the inverted term and stem indexes in memory helped to achieve faster calculations.
- BM25 was found to provide better accuracy and more documents ranked when run on cranfield data index.
- Lemmatization of valid tokens and storage. Implemented dictionary compression using Front Coding.
- Postings lists and specific document statistics were compressed using Gamma and Delta compressions respectively.
- Reduction of approximately 65% in full index size was evident after compression for the Cranfield dataset.
- Worked in a team of 3, to develop a recipe sharing/ blogging app using Java technologies and Agile UP principles.
- Implementation of Adapter pattern, Factory pattern and Singleton pattern according to strict OOP principles.
- Project was developed in 3 iterations following Agile based Object Oriented development patterns. Technologies used were Java, JSP/Servlet, MySQL.
- Developed a web-application to manage and display flights, passenger details, booking information with simple database search feature.
- Implemented server-side module classes for connectivity, CRUD and join operations, for MySQL using MySQL Connector.
- Increased waiting time for database operations was countered by loading required data into in-memory Hashmaps
- Ruby on Rails based web-application to buy and sell used cars, along with Craigslist like abilities and interface using Agile methodology, in a duration of 3 weeks. Followed integration testing approach, and developed in a team of 2.
- The task was to develop a website fingerprinting counter attacking analyzer.
- Implementation of classification techniques to provide a temporally improving model for better online-attack analysis
- We achieved an average 70% accuracy at best using our Ensemble of classifications