Portfolio - UBS

http://www.ubs.com/

At UBS my primary responsibility was building a system to handle risk analysis of bond backed CDO trades.  The risk analysis was developed by in-house quants.  Our data was sourced from a variety of feed services, Excel files, databases, and the traders themselves.  Given the amount of data we had to handle, we built separate OLAP and OLTP databases.  

As we received data, it was loaded into the OLTP database.  Only current production information was stored here.  Whenever the daily production runs were ready, the data would be moved into the OLAP snowflake schema.  The aggregated data would then be extracted from the OLAP database, converted into XML, then sent off to distributed machines to run the risk analysis.  Once risk analysis was done, data was correlated back to the snowflake schema and reports were generated.

Traders were also able to schedule bespoke reports for their own scenarios.  They would define some alteration to the data (replacing a bond, skewing the risk prediction curves, ultimately any change conceivable).  This change would be loaded into the snowflake schema as a branch to the existing data.  This allowed traders to get day-over-day reports of their customized deals.

During this project I was the core designer and lead developer.  The team varied from 10 to 15 developers along with other support staff.  The project was completed ahead of predictions and boasted a robust enough architecture to allow it to be drastically repurposed nearly immediately after launch to analyze risk of other complex instruments.
Comments