Mar 2019 – Jan 2022 · 2 yrs 10 mos · Cambridge, United Kingdom

• At Samsung Research, we were developing innovative Artificial Intelligence (AI) technologies to enhance current business and open up new business opportunities, with the ultimate goal of enhancing human life and contributing to the society.
• During my period in Samsung I proposed, lead and worked in several ML initiatives, many of these resulted in publications in top-tier peer-reviewed conferences, US patents, and even dedicated engineering teams to deliver ML products. On top of my work at Samsung I have also organized international workshops and tutorials as well as participated as a reviewer and speaker in several top-tier conferences. Among several top-tier conference papers, in 2021 our work "Fjord: Fair and accurate federated learning under heterogeneous targets with ordered dropout" was accepted to Neurips, arguably the most reputable international ML conference as a spotlight paper (i.e., top 3% of all papers accepted).

Jul 2018 – Mar 2019 · 8 mos · London Area, United Kingdom

• Mosaic’s mission is to increase transparency, understanding, and trust in the cryptofinance ecosystem by enabling retail and institutional clients to access and build applications on our data platform as well as access to deep qualitative and quantitative research. Partners include large financial service providers, hedge funds, venture capitalists, and leading cryptocurrency wallets and exchanges.
• I was the first developer to be hired and had a significant impact in building the team and designing the overall architecture. I was responsible for ingesting data from hundreds of exchanges and dozens of blockchain nodes, setting up the data pipelines, databases and lastly all the monitoring and analytics. As I progressed through the stack also hired and mentored other engineers to extend the platform. I was also responsible for several investor and client meetings, in the latter I translated their requirements into actionable items for the development team.
• The main challenges had to do with the real time properties of the system, such as on-the-fly etl, redundancy of streams, resilience to API changes, on-the-fly analytics and efficient database operations.
• Technologies : terraform, ansible, kinesis firehose, s3, redshift, kubernetes, elastic search, lambda, ec2, python, JS, influxdb, grafana, etc

Sep 2017 – Dec 2019 · 2 yrs 3 mos · Delaware, United States

• Tappas mission was to make mobile application (app) testing as easy, effective, and ubiquitous as possible. Tappas augments the scale of interactive testing via planetary scale crowdsourcing and autonomous AI tools trained by real humans.
• I was one of Tappas founders and was responsible for building the engineering team and designing the architecture, as well as conducting investor meetings.
• Tappas architecture was highly scalable, virtualizing mobile OSes in the cloud in an elastic, demand-driven fashion. It automated the full process of launching VMs in real-time, pre-loaded with the customers apps and streaming the application to real human testers across the world. On top of this, Tappas provided an extensive analytics web platform, showing videos of the testers interactions, heatmaps, tasks, customizable feedback queries and advanced low-level metrics such as per functionality latency and code coverage. To completely distance itself from the competition, Tappas also used state-of-the-art recurrent neural networks, to mimic user interaction in ways that current automation tools cannot.
• Overall it presented several challenges when scaling to thousands of testers and apps, from fast VM launching with copy-on-write and storage in-memory, to sharing the communication channel between streaming and fetching statistics (e.g., tcpdump logs, videos, click information, call graph and code statistics, etc) to assure good QoE and at the same time avoid nodes getting overloaded, and scaling on-demand to keep costs low.
• The stack consisted of: Terraform, Ansible, Qemu, VNC, tcpdump & other network tools, Elixir, Phoenix, JS & several visualization libs, Grafana, Elastic Search, PostgreSQL, InfluxDB, AWS, Pytorch, etc

Jan 2016 – Jan 2017 · 1 yr

• My project inside Telefonica was awarded funding to incubate a startup with a team of business, marketing and developers. I lead a team of developers that aimed to release mobile phone proxys consisting of a bare phone with a screen and antenna that would serve as a portal to a fully virtualized mobile operating system executing on the cloud or on the ISP premises (e.g. near the cell phone antennas).
• As part of the experience I had to delegate work and conduct meetings with partners and potential clients.

Sep 2013 – Aug 2017 · 3 yrs 11 mos

• Telefonica is the number one Spanish multinational by market capitalisation and
• one of the largest private telecommunications companies in the world. Its
• research department focuses on global, technical innovation trends.

Jan 2013 – Aug 2013 · 7 mos

• Master thesis @ Telefonica research department.

Sep 2015 – Dec 2015 · 3 mos · Greater Cambridge Area

• Partitioning and distributing mobile apps in the network to extend battery life time and speed-up execution. Algorithms for computation distribution in the NFV paradigm.

Sep 2013 – Aug 2017 · 3 yrs 11 mos

Aug 2012 – Jan 2013 · 5 mos · Greater Stockholm Metropolitan Area

• We implemented the first truly highly-available and scalable Hadoop YARN solution that aimed to allow the existence of multiple stateless Resource Managers using the NDB MySQL Cluster.