Quantitative Analysis
Physics and statistics, joined directly to a decision.
I was a physicist at Bell Laboratories and was recruited into the industry in 1986 as a quant supporting proprietary trading, which is how that generation arrived. The quantitative work spans derivatives and fixed-income modelling, portfolio construction, risk and performance attribution, and applied statistical learning — and it has almost always been attached to a trading, investment, or risk decision rather than left as an exercise.
Bell Labs, and the record with Jan Dash
Jan Dash's Quantitative Finance and Risk Management: A Physicist's Approach (World Scientific) names me at several points: for bringing Prony analysis to his attention at Bell Laboratories, for the question that led him to connect stochastic equations and path integrals in finance, and for ideas on a firm-wide economic-capital utility function.
The mathematics of path integrals in finance is Jan Dash's; the field credits him correctly, and so do I. What the record shows is narrower and, to me, more interesting: the question came out of a conversation, and I was in it. I founded the quantitative research effort at Merrill Lynch and recruited Jan Dash from Bell Laboratories to deepen the mathematical work.
I have known Santa since we were both at Bell Labs in the 1980's and both went up to the Street. He has a distinguished and unusual career with expertise in risk management, trading, portfolio management, and quant issues — fluently speaking many languages of finance professionals. Santa provides deep insight with practical analysis and advice. I recommend Santa without hesitation.Jan Dash, PhD · author, Quantitative Finance and Risk Management
Applied quantitative work
Derivatives and structured-product valuation; fixed-income and mortgage modelling; portfolio construction and optimization; risk and performance attribution that separates genuine alpha from beta; VaR and CVaR; and a set of methods for dependence, concentration, and early warning — copula tail dependence, Hawkes processes, factor and network models, PCA, entropy concentration, community detection, and Bayesian updating. At Perry Capital this became a risk-factor and attribution framework and a profitable, actively traded macro overlay that controlled risk and enhanced returns rather than only constraining them.
Fund and manager assessment
A distinct line of the work, and one I have been asked for repeatedly: applying performance and risk attribution to a fund business rather than to a trading book — separating genuine alpha from factor exposure across a book of hedge-fund investments, and advising on manager, strategy and portfolio-construction choices on that evidence. I built the quantitative framework for exactly that at Cowen, having been brought in by the head of the investment division; I built the attribution framework that distinguished skill from beta at Perry Capital; and I founded a completion fund-of-funds on quantitative and fundamental manager selection with hedge-fund replication.
It is worth saying how that work tends to arrive. Almost every time, it has come from someone who had already seen the work — a fund investor who had watched me build a risk function, a chief financial officer who hired me twice at two different institutions, a firm that hired me twice thirteen years apart. That is the only reference I would offer for it.
Published
Four chapters in Credit Derivative Strategies: New Thinking on Managing Risk and Return, edited by Rohan Douglas, Bloomberg Press, 2007 — risk management of credit derivatives, four synthetic-CDO trading strategies, CDS valuation, and CDO valuation. Written at the peak of the credit-derivatives era, on the instruments that subsequently defined the crisis. I lectured on the same material at the NYU Courant Institute.
Still live
The physics is not only history: a 3+1-dimensional computational study of the graceful-exit problem in inflationary cosmology, with a numerical percolation-threshold result, running back to the Princeton thesis.
Boundary. Applied statistical and machine-learning methods — logistic regression, PCA and clustering, community detection, anomaly and regime detection, statistical pattern recognition — yes. Deep-learning engineering, frontier-model training, and MLOps infrastructure: no. I do not claim them.
The systems these methods run inside → · The published record and citations →
A difficult modelling or valuation question.
For derivatives and portfolio analytics, risk attribution, model review, and early-warning or event-consequence modelling.