Robin Tyagi currently heads the Data Science function at Vivriti Capital. In his current role; he has spearheaded loss modelling frameworks for complex products (securitization, direct assignment) under retails and balance sheet level credit scorecards and guarantee backed products (for wholesale loans) to ensure high levels of accuracy and has helped maintaining near zero delinquencies on the portfolio.
Robin has close to a decade of work experience and has worked upon multiple facets of credit risk modelling for financial institutions - development of economic capital estimation models/expected credit loss models to estimate losses against credit defaults on portfolio and event risks for wholesale & retail balance sheet, estimation of losses and probability of default using transition state matrix based on historical portfolios , development of statistical & machine learning models to under borrower credit behaviour in various asset classes.
Additionally, Robin has worked closely with Structured Products team to gain expertise in variety of debt capital market products including - term loans, rated securitization, bilateral assignment, non-convertible debentures, guarantee-backed financing, commercial paper, long-term subordinated debt instruments and preference shares across different asset class for entire product lifecycle. Additionally; he has also developed loss estimation model & rating framework for debt AIFs- which has been used by CRISIL to rate these instruments.
In his earlier role at CRISIL; he has also worked on variety of equity OTC derivatives product booking & payoff reconciliation like Snowballing Auto-call Notes, Binary/Digital Notes, Rainbow Option, Twin Win, Look Back, Range Accrual Notes, Barrier Options, Asian Options, Packaging notes, Customized Parsers etc; where he liaised with various internal stakeholders like traders, trade support teams, documentation teams, Quants Teams, Product Valuations Group, P&L teams etc. to sort out discrepancies between trade booking & documentation and build/test pricing models for these products