Impairment Report
Problem Statement
A British multinational investment bank has more than 2000 RMBS Cusips, that are valuated every month.
An excel-based Risk Model is used for computing Prepay/Default/Severity vectors.
These vectors are manually passed to compute Cash-Flow and Perform the Required Valuations.
Manual process that requires a month to complete analysis over a month’s remittance cycle.
Accuracy and Analysis got affected due to the Manual Process, requiring automation to speed up analysis and improve accuracy.
Solution Implemented
A Web-Portal to set up User Defined Portfolios and specify Run Parameters like interest rates, HPI scenario, paths.
A CPP-based Replica of Risk Model Adapter for Computing Vectors.
CashFlow Wrapper Based on Intex ICMO API to consume vectors and create loan-level and bond-level cashflows.
Amazon Web Services-based Cloud Infrastructure.
Excel-based Impairment Report with Tabular and Graphical Analysis of the Projections.
Grid Controller to utilize the Computing Power of Multiple Processors simultaneously.
Processing time for 2000 RMBS bonds reduced to 4-5 minutes with 100% accuracy.
Bond Analytics Platform
Problem Statement
CA-based Housing Data Analytics corporation needed a platform for bond analytics and credit reporting.
Used Loan Performance Model to generate Prepay Default vectors and Intex ICMO API to generate cashflow.
Needed a Web-Based Bond Analytics Platform for defining HPI / Interest rate scenarios, performing valuations, and accessing Collateral Stratifications.
Solution Implemented
Loan Performance and Intex Data Mart with a Web-Based Platform for HPI/Interest Rate scenarios.
CPP engine utilizing Intex ICMO API and Loan Performance Risk Model.
Interest Rate Simulator using QuantLib for Forward Rates.
Excel-based Credit Report with detailed Price/Yield Projections, Adverse Stratification, and Delinquency Analysis.
Grid controller built on Java Space for utilizing Multiple Computing Resources.
RMBS Portfolio Valuation
Problem Statement
Bond analytics Platform (BAP) was expected to be used to evaluate Portfolio on a monthly basis for a number of Scenarios.
Portfolio has more than 40,000 cusips. Analysis of each cusip was supposed to be performed over 200 simulated paths of Interest Rate and HPI Scenarios.
Solution Implemented
Augmented BAP with a specialized process to use HJM Model to simulate Interest Rate, HPI and Generate Risk Vectors for each path.
Deployed BAP on Amazon - Cloud Environment.
Developed an automated process to Synchronize Amazon cloud setup with internal Production setup, used computing powers of more than 100 nodes simultaneously.
Developed a Cash-Flow consolidator Application, that combined CashFlow for each Cusip to get one view analysis for the whole Portfolio.
OLAP Cube
Problem Statement
Needed rapid access to ADHOC multidimensional stratifications of Loan Level Data.
A Loan Universe of more than 20 million Loans that receives updates on Daily basis. A Process needed to manage daily updates.
Stratification ranging from simple FICO/LTV buckets, Property types to more complex ones like delinquencies and default behavior.
Also, Historical Data of a Decade for entire Loan Universe was supposed to be Updated.
Solution Implemented
Developed a Relational Database and an ETL process to load historical loan data.
An Intex ICMO API based CPP Application to extract strip ratio for each CUSIP.
Loan Level, Deal Level and Pool Level Mapping between Intex and Loan Performance.
Developed an OLAP cube using SSAS that had all Measures, Dimensions and Bucketing. Allowed Loan Level drill-down.
ASP .Net Web Client for accessing the OLAP cube through web-portal.
Excel based credit reporting and loan level reports.
Whole Loan Pricing
Problem Statement
Time-sharing services company that later became known for providing financial market data, analytics, and related solutions to financial institutions, active traders, and individual investors, has a large portfolio of unsecuritized loans that it wants to valuate every day.
Tens of thousands of tranches get evaluated covering over 5,000 deals, that has to be used as reference for the pricing.
In addition to this, ADCo model, and a regression technique to derive a spread function that relates various loan characteristics to a yield spread should be used for pricing.
Solution Implemented
An Intex ICMO API-based engine that uses tranche prices and group level vectors to calculate discount margin for each tranche.
A regression engine runs over key features of loan portfolio (like FICO, LTV, property type, occupancy status, Current/original balance etc.) and fits in a discount margin to each loan.
Developed a risk model adapter using ADCo Model to predict prepay/default characteristics for unsecuritized loans and generate loan's cashflows, and finally discount the cash-flow to come up with price of each loan.
Trustee Data
Problem Statement
United Kingdom's largest property group, including residential property surveying, a collaboration of estate agents, and corporate services, has a Risk Model that is driven by Loan Level secure Data.
This data is made available on respective trustee websites ahead of data vendors.
As an advantage of this, data was to be extracted from trustee reports as soon as it gets updated on website and run model.
Solution Implemented
Developed a Java-based website crawler to search and download reports from different websites.
Implemented a robust monarch model for each trustee to extract loan level and deal level data from trustee files.
A relational database schema and an ETL process to upload this data into database, generate input files and run risk model.
A validation mechanism to flag data wherever significant difference from previous month was observed.
An automated script to run all this in case of a new file uploaded by trustee.
Ticker Plant
Problem Statement
A large scale ticker plant was needed for Indian market.
Some of the Major requirements were as follows - Process and Distribute Real-time Data from various Indian stock and commodity exchanges, viz. NSE, BSE, MCX, NCDEX, NMCE.
Process and Distribute News from multiple sources like CRISIL and IFW, Contribution engine to publish exchange rate quotes, publish news, grey market quotes etc.
An in-built redundancy and ability to delay quotes in terms of Architecture.
Solution Implemented
CPP based feed handlers as windows service to connect to stock exchanges, parse data, and publish it to streaming client.
Calculation engine (CPP windows service) to perform live and SOD/EOD calculations, send them to streaming client, also handled option calculations, Greek Calculations, any corporate action adjustments. Real-time ranking engine to publish top gainers, losers etc.
News handlers, Excel based contribution engine that publishes quotes/news to server.
Relational database to store quotes, news, calculated fields. A number of stored procedures to perform complex calculations.
.Net based ticker plant admin to configure applications.