Lisbon, Lisbon, Portugal
Building + scaling up high performance systems and teams from scratch, with the mission of providing reliable and lightning fast global remittances for millions a day. We help migrants move hundreds of millions EUR a day to their loved ones across dozens of countries 🌍❤
Building a reliable, fast and cost-effective global payments platform from scratch, specialized in helping migrants send money back to their loved ones.
Part of the Payments team, and thus focused on making pay-in, pay-out, transfer execution, and pricing mechanisms more efficient, reliable and scalable, including SRE duties (on-call, incident response).
I led a team which focused on research projects for Routing, supporting Huawei's Petal Maps. These projects varied in nature but generally sought to introduce novel algorithms and machine learning models which would improve the quality of service for routing, whether in terms of performance, accuracy, user experience optimization or cost efficiency. Even though these were research projects and we'd work primarily with prototypes, we treated code quality very seriously and thus also developed & maintained the whole devops infrastructure required to quickly deploy and test our prototypes. Some examples of the work I did: - Introduced speed-up techniques to state-of-the-art CRP algorithms - Developed ultra-accurate and personalized ETA models through machine learning - Developed two key projects for worldwide use, in Huawei phones
- Development and maintenance of routing services supporting Daimler autonomous vehicle operations, in specific for San Jose pilot (node.js, javascript, gitlab, docker, kubernetes, apache kafka) - Design and development of automated routing graph "masking" solution for constraining base map to operational design domain, as defined by multiple HD-map providers (python, map matching) - Development of synthetic routing graph generation framework for testing of graph "masking" solution (python) - On-call service supporting San Jose pilot operations
I did R&D for Routing - in particular, I developed methods, models and software which make use of big and small data in order to improve our products. This usually entails a great deal of big data engineering, data wrangling, ETL, data exploration, predictive modeling via machine learning, prototyping, hacking/gluing of complex systems, and also presenting results in an insightful manner, with minimal visual complexity. For some of the most promising prototypes and models, I would carry on the work all the way to production, by being involved on all stages of the development process: system design and architecture, development, testing and deployment (often using cloud services and virtualization technologies). The last example of this was the development a route matching algorithm based on a Hidden-Markov Model which is available as a native C++ library and an internal REST API for benefit of mine and other teams.
In this role, I focused on the quality of route choice, travel times and other sub-products of HERE Routing, which are of non-trivial assessment. My team's ultimate goal is more tied to applied research: typically it involves designing and developing tools and methods to assess the status quo, driving improvements by devising new model experiments, recommending those that work to our product teams and finally collaborating with our engineering teams in order to implement them. An example of such activity is the assessment of the accuracy of travel times produced by our products against ground truth data, for which I built a scalable diagnostics framework based on Spark and Python. This framework has been used to decompose and understand travel time error and eventually to experiment new models and provide concrete system recommendations across our software/content stack.