I have been designing and developing web pages for about a decade now. I have detailed knowledge in coding practices, creating clean and usable code along with simple yet beautiful designs. In the past few years I have focused more on the backend.
Most of my time is spent running and developing an ad serving platform written in PHP and Python. We serve around 350 million PI per month. The 99.9% percentile is under 20ms. Data is king, so we analyze every single interaction our users have with our system. We're able to continuously optimize the way our banners look, what banners are served to specific users, and which campaigns perform best for certain geo locations. The analytics platform (built with Django) allows us and our customers to track all necessary metrics in realtime. We use map/reduce for more in-depth analytics and fraud protection. Our real time bidder (RTB) takes part in tiny auctions that sell single ad impressions. This service is written in Python using evented I/O and relies on a hand-written Redis cluster system on EC2. 99.9% of requests to the data cluster return in less than 1ms.
Before, I have worked on main-memory databases at SAP Labs in Palo Alto, CA. Some of them are row-oriented, some happen to be column-oriented. We researched what happens if column-stores are used for OLTP-style transaction-based processing. We also looked into hybrid data stores that leverage the performance benefits of both architectures, depending on how the database is actually used and which data is stored.
Database technology is one of my primary interests in computer science. I parse, archive, analyze and visualize just about every bit of data I can get my hands on. A while back I developed an intelligent sharding mechanism with exploits the declarative nature of Django's data modelling. We planned on building a scalable social network based on Django. From day one it would have to handle the millions of requests Foren-City handles every day. With sharded Django models we were able to benefit from the Django models architecture while running on a sharded MySQL cluster.
Data mining, natural language processing, and collective intelligence are some of my primary interests along with database systems. Machine learning and database courses in college led to a music recommendation site that used Latent Semantic Indexing and K-Means (muxfind.com), and a project on effective agglomerative clustering. I have co-authored a paper about large-scale storage and aggregation of RFID data. For my bachelor's thesis I did research on large-scale main memory databases in cooperation with SAP.
I learned most of what I do through trial & error, with a hint of college courses and relevant textbooks.