Stripe Data Consultant2019 - PRESENTAbundance Digital
Technologies: Stripe Sigma, Python, Ruby on Rails
- Answered relevant business questions based on available Stripe data for this community-based company. The answered resembled their subscription model and how well does it perform.
- Implemented a checkout page to allow people to pay for the yearly subscription on a monthly basis using Splitit.com.
- Implemented relevant SQL Queries in Stripe Sigma so that the business owners can always run the queries on the latest data.
- Helped to automate their boarding process with Zapier.
- Extensively analyzed the subscriptions data and suggested ways to improve churn rate, revenue and profit.
Software Engineer Intern2019 - 2019
Technologies: Python, C++
- Researched new approaches to build a request predictor - a time-series discrete value model.
- The problem resembles Intermittent Demand Forecasting. The approaches considered include Croston's method, Bootstrapping, RNNs and.
- Implemented and tested the performance of different algorithms on previously acquired data in Python Colab.
- Implemented the method that performed best in our dataset.
- Proactively proposed different methods to make the project a success.
Software Engineer Intern2018 - 2018
Technologies: C++, Protocol Buffers, R, Machine Learning, Unix
- Built a fully automated machine learning model training and prediction pipeline. This was the first step in moving from alpha testing where we get the predictions manually by running R scripts, to a beta phase.
- Ensured the data generated by the pipeline is consumed by an internal dashboard. A strong motivating factor for building an automated pipeline is that the data is always updated and will solve the out-of-sync data problem caused by communication delay between the Engineering, Product and Sales teams.
Full Stack Developer2018 - 2018One Step Software
Technologies: Ruby on Rails, Heroku, React Native,
- Worked as a full-stack engineer for a sober-living SaaS platform.
- Built the CI pipeline in order to make sure our service did not introduce any bugs.
- Collaborated with the team to make our Trello better and advocated for a TDD approach in order to make sure we serve the best software to their clients.
- Contributed to the Rails web app, introducing new features, implement backlog tasks, and fixing existing bugs.
- Developed the API for the React Native mobile app and then used that API to build the mobile app.
Back-end Engineer2017 - 2018Kuende
Technologies: Scala, GoLang, Ruby on Rails, gRPC
- Contributed to microservices for interacting with Google Cloud storage and Google Cloud PubSub.
- Developed a GoLang gRPC server that interacts with Google Storage.
- Built a Scala gRPC client that connects to the Go server. The reason for doing this is that Google Storage SDK does not currently support Scala.
- Wrote unit, integration, and end to end tests.
Full stack web developer2017 - 2017NLV Production
Technologies: Ruby on Rails, Heroku, PostgreSQL, Sidekiq, Stripe, MailGun
- Contributed to the implementation of new features and fixing bugs for the nlvproductions.com website.
- Used Stripe as a payment processor and investigated the problems they were experiencing with customers being charged more.
- Implemented a new workflow for the admin dashboard that improved the organization's processes.
- Been very proactive and provided multiple solutions for every problem that the customer was interested in solving.
- Fixed memory leaks we experienced on the Heroku dyno.
Software Engineer Intern2017 - 2017
Technologies: Java 8, Protocol Buffers
- Interned in the Google Photos Team. The project involved extending the selection framework for auto-created movies based on user content and implementing a few new concept movies.
- Created the concept "child friendly" movie. The basic idea is that if we detect that a lot of pictures contain kids, toddlers, or children, we select music that is suitable for a child movie.
- Designed a second concept movie called 'Outdoor Movie' and it basically tries to match outdoor pictures with suitable outdoor songs.
- Conceptualized a third movie called 'Highlight Reel' and the main idea is that we want to crop the user's videos so that we retain only some 'action moments'. An action moment is a 5s clip in the movie where we detect some action (i.e. throwing a basketball or shooting a baseball).
Mobile Developer2016 - 2017Kuende
- Contributed to the team that develops the Kuende mobile application (available to download on AppStore and Play Store), a social network based on gamification.
- Implemented a Trie based algorithm for searching people, a wrapper around a list-view so you can easily declare, add filters and comparators to lists, and other small features.
- Optimized the way we persist our data so that the information is fast accessible and rendered on the phone.
- Concepted and implemented the Flux architecture on top of C++ and Qt.
Teacher2016 - 2016LearnHouse
Technologies: Advanced algorithms and data structures in C/C++
- Taught computer science fundamentals to high-school students through the non-profit organization (http://learnhouse.ro).
- Presented advanced algorithms and data structures to students interested in the National Olympiad in Informatics.
- You can find some of the problems that I solved on various competitive programming challenges can be found here: https://github.com/rusucosmin/cplusplus.
Software Engineer Intern2016 - 2016Amazon
- Interned in the Robot Detection Research Team at Amazon Development Center Iasi.
- Created a linkage and score system for robotic attributes. The main challenge was the huge amount of nodes and data that had to be processed every day.
- Contributed to mitigation of all of the robotic activity that crawls our website, with two things in mind. First, we didn't want to block a real user! Second, we wanted to let some friendly bots crawl the website (eg GoogleBot, BingBot).
- Designed a system that was able to give a score for each incoming request on the website. Then, based on that score we can decide if the current request was made by a human or a robot. For this approach, we split the request into simple attributes such as user agent, IP, customer id and we independently compute the scores for each such attributes. In the end, we can average the scores to get the request score.
- Computed the scores as a series of Hadoop Map Reduce jobs that ran on Amazon's Hadoop Clusters and infrastructure. I had a lot of freedom with the project and I learned a lot while designing such a scalable, distributed algorithm.
Android Developer2013 - 2013e-communio
- Built the e-communio app based on the already built local church news website.
- Understood the main concepts of Android OS and the process of publishing an app.
- Created the app as a very simple, and at the core, just a list view, containing the latest news that I parsed from a web service written in C# that connects to the database, gathers the news, and splits a JSON array.
- On the mobile side, I created a request for the news, parsed them and displayed them. When the user clicked one of the items, it will go to another screen where they could read the entire content of the article. There were other nice features such as increasing and decreasing the size of the text, sharing the article, and additional text effects.