Computer Science Assistant

Andrei V. Anghelescu

** **** ********** *****

Princeton, NJ 08540-6314

******@**********.**

Summary

Highly motivated trading desk strategist with two years of experience in the MBS and ABS areas.

Experienced with development of models of nancial products, as well as coordination of design and

implementation in conjunction with the trading desk. Computer Science Ph.D. level analytical back-

ground with expertise in Machine Learning, Discrete Optimisation, Information Retrieval, Structural

Data Analysis and Visualisation, and their applications in text categorization, bioinformatics and

time series analysis.

Skills

Expertise in a variety of machine learning, statistical analysis, optimisation techniques and

numerical methods.

Business knowledge of diverse securitized products (ABS, MBS, CMO), rate and loss models

(HPA, prepay, default). Ability to quickly learn new products and domains.

Expert in C++, Python, Perl, various UNIX shells and Java. Long-standing experience with

various libraries: STL (Standard Template Library), Boost, Common C++, Lemur (text

retrieval), ACE (Adaptive Communication Environment), Matlab and Zope.

Focus on design ergonomy and minimalism; able to represent tasks in exible abstract concepts.

Strong written and interpersonal skills, with several years of experience teaching and present-

ing. Comfortable and e ective speaker in all situations.

Fast integration into a team; e cient when working individually.

Fluent in English, French and Romanian; pro cient in German. International experience

obtained from working in three di erent countries.

Permanent resident of the USA (green card holder); EU citizen.

Professional Experience

Desk Strategist Morgan Stanley

2006-present New York, NY

Responsible for modelling and risk management development for the primary market trading desk,

and for the new products originated in the retail banking group (MS Bank). Developed, selected

and a rmed the models used by the desk as a rst step in the model certi cation process. Expe-

rience working on residential mortgage trading desks (primary, secondary) on the development of a

variety of models and tools. Designed the infrastructure for the client mortgage rate model. Long

term collaboration with the asset backed securities desk towards the development of custom models

required by the traders. Developed a prototype of a blotter for the RPX trading desk. Consistently

interacted with the traders regarding strategy design and development, rapidly adjusting focus and

direction as prompted.

Project highlights:

contributed to the development of an analytics library that modelled the various structured

products traded by the aforementioned desks. This included mortgage pool optimisation,

pricing of mortgage securities, servicing rights and custom loans, and mortgage pipeline man-

agement models.

developed analytical tools for risk valuation and reporting for various custom products origi-

nating in the retail banking division.

designed the infrastructure for the deployment of residential mortgage rates to clients, then

supervised its development.

contributed to the design and development of an application that integrated all securities of

various desks and databases. Interfaced with analytics libraries to allow seamless computation

of metrics and risk management. Working prototype of data caching to allow live update of

metrics where computationally feasible.

in collaboration with the modelling team developed an application for computing mortgage

servicing rights - initially with myself as sole contributor, later as a supervisor of the imple-

mentation of this model as a distributed application that enabled heavy computations.

developed a simple default model for HELOCs and calibrated it to recent market prices - the

current market developments were both the chief reason and the main challenge.

developed a variety of frequently used applications to streamline the daily tasks of the desk

(pricing, marking, risk management). Many of these applications were later included as com-

ponents of core tools.

Graduate Intern Pharmacopeia, Drug Discovery Inc.

2004-2005 New Jersey,USA

Created a patented technology for estimating molecular 3D structures using multiple alignments of

1D representations of molecules. Developed a scalable clustering procedure for molecular ngerprint

data. The scalability was a key factor, as the size of a molecular library is much larger than what

can be handled by state-of-the-art hardware. The software was written in C++ as a library with

a small text-based interface, and it added 15,000 lines of code to the existing software, on which it

relied.

Patent Application Nr. 11/283650

Graduate Assistant Rutgers University

1999-2006 New Jersey, USA

Research and teaching assistantships in DIMACS (Center for Discrete Mathematics and Theoretical

Computer Science) and the Department of Computer Science:

Research assistant in the Knowledge Discovery and Data Mining (KDD) sponsored Moni-

toring Message Streams project at DIMACS. Wrote (in C++) an object-oriented library of

classi cation and feature selection algorithms, data representation, wrappers to SVM learning

software (i.e. SVMLight ) and other machine learning related algorithms. Published under the

GNU Public Licence v2.0. The implementation focuses equally on speed of execution, small

memory footprint and convenience of use for the programmer. Currently at 22,000 lines of

code, available at http://www.cs.rutgers.edu/~angheles.

Research assistant in the Computer Science department at Rutgers University. Development

of novel clustering methods exempli ed by applications in text analysis, bio-informatics and

nancial time-series data.

collaboration with the Linguistics department at Rutgers, in the development of a study of

anaphora in African languages.

Teaching assistant in the Computer Science Department, Rutgers University. Courses taught

include: Numerical Analysis, Internet Technology, Data Structures, Computer Graphics.

Software engineer Bouygues Telecom

1998-1999 Paris, France

Various projects in the Network and IT department.

Developed automatic cell frequency allocation software, using Monte Carlo methods (Simulated

Annealing and Genetic Algorithms). Integrated with PlaNET R and CellOpt R .

Supervised and contributed to the development of a hardware inventory distributed database,

with Oracle R back-end and Delphi-based front-end.

Systems administrator University of Bucharest

1997 Bucharest, Romania

Maintained Windows NT network. Updated institutional database. Assisted in the roll out of

various software programs.

Programmer Sumus S.R.L

1996-1998 Bucharest, Romania

Developed and maintained various programs with applications in civil engineering.

Education

2009 Ph.D., Department of Computer Science, Rutgers University (expected 2009)

1998 B.S., Department of Mathematics, University of Bucharest

Selected Publications

[1] Andrei V. Anghelescu, Robert K. DeLisle, Je rey F. Lowrie, Anthony Klon, David J. Diller. Tech-

nique for Generating Three-Dimensional Alignments of Multiple Ligands from One-Dimensional

Alignments. In Journal of Chemical Information and Modeling, vol. 48-5, 2008.

[2] Ken Sa r, Andrei V. Anghelescu, Sarah E. Murray, Jessica Rett. The African Anaphora Project.

In Proceedings of the Language Resource and Evaluation LREC Conference, 2006.

[3] Andrei V. Anghelescu, Aynur Dayanik, Dmitriy Fradkin, Alex Genkin, Paul Kantor, David

Lewis, David Madigan, Ilya Muchnik, Fred Roberts. Simulated Entity Resolution by Diverse

Means. In Proceedings of the KDD Challenge Cup, 2005.

[4] Andrei V. Anghelescu and Ilya B. Muchnik. Optimisation of SVM in the space of two parameters:

weak margin and intercept. Applications to text classi cation. JICRD, 2004.

[5] Andrei V. Anghelescu and Ilya B. Muchnik. Combinatorial PCA and SVM methods for fea-

ture selection in learning classi cation (applications to text categorisation). In Proceedings of

the IEEE International Conference on Integration of Knowledge Intensive Multi-Agent Systems

(KIMAS 03), pages 491 496, 2003.

[6] Andrei V. Anghelescu, Endre Boros, Dmitriy Fradkin, David D. Lewis, Vladimir Menkov,

David J. Neu, Kwong Bor Ng, and Paul B. Kantor. Prospective data fusion for batch lter-

ing. JICRD, 2003.

[7] Andrei V. Anghelescu, Endre Boros, David D. Lewis, Vladimir Menkov, David J. Neu, and Paul

Kantor. Rutgers ltering work at TREC 2002: Adaptive and batch. In Proceedings of the 11th

Text Retrieval Conference (TREC 2002), 2002.

[8] Andrei V. Anghelescu, Ilya B. Muchnik, and Casimir A. Kulikowski. Categorization of scien-

ti c papers by consensus clustering from pre-de ned lists of keywords: Application to medical

informatics. In Proceedings of the Medical Informatics Symposium in Taiwan, 2002.

[9] Casimir A. Kulikowski, Ilya B. Muchnik, Akshay Vashist, Andrei V. Anghelescu, Hwaseob J.

Yun, Eric Linton, and Joachim Messing. Multi-alignment of paralogs for functional annotation:

Application to the rice genome. In Proceedings of the 5th Annual Conf. on Computational

Genomics, Baltimore, 18, 2001.

Contact this candidate