Hi, I am a diligent and passionate Software Engineer, particularly interested in backend, data science and data engineering and applied computer Science , currently working as a Software Development Engineer, intern at Rivigo. Prior to joining Rivigo, I was an undergraduate student at Indian Institute of Technology, Kanpur, Department of Computer Science and Engineering. During my BTech days, I interned in American Express, Big Data Labs, Bangalore under the mentorship of Mr. Rahul Ghosh. Curious to work in a startup, I completed my B. Tech in seven semesters and joined Rivigo(as opposed to Microsoft research, Bangalore) as an intern since January 2019. In Rivigo, I have been actively working in the Data Science team over a series of projects in pricing and goal app.
Currently, I have a job offer from Goldman Sachs and am all set to join there and prove my worth on May 27, 2019. As a part of Rivigo, I have enthusiastically worked and tried to learn as much as I can. Though it has been only three months for me in Rivigo, I feel that this has brought a tremendous change in myself, especially the difference in my coding style, six months back. I also realized the importance of Systematism in the way you code and the use of IDE’s like Pycharm, task management tools like Jira, confluence, etc, and above all, a more effective way of using git as a VCS. Other great open source tools I came across include superset, zeppelin and metabase and some good python libraries like pandas, pymongo and sqlalchemy.
In this period, I have worked on data collection and feature preparation for the operations team. This work of mine tries to predict the ODVT prices that need to be recollected based on the price confidence scores. Due to some delays due feature improvements, I switched over to prediction in prices, where I learnt and appreciated the greatness of Pandas and Jupyter Notebook. This work lasted roughly for about 20 days or so, where I tried to compute ratio of trip prices in weekend and in weekday, for every ODVT, doing a hierarchical substitution by testing the required hypothesis. If the hypothesis turned out true, we then computed the ratios for them. In case the hypothesis turned out False we tried for higher levels, like family and all trucks, and then tried to analyse whether the factor multiplication over routine prices holds good or not. Apart from this, I worked on utilizing SQLAlchemy for optimizing the SQL queries as well. While using it, I found a documentation bug there as well, which then I reported as an issue and made a PR for the same. PS the PR is yet to be accepted.
Another and one of the most interesting things I came across here was clipper, a low latency online prediction serving system, developed by students of UC Berkeley and UChicago, which I checked out and tried to observe its features and also panelled A snapshot of some of the demo code snippets which I illustrated can be found on my github.