Just returned from an 8 hours torture. Thanks for the helpful messages to my last post.Now I have other two superdays coming up, one is for model validation position, one is for fixed income quant. I would very much apprecaite any insight on typical questions for such roles and any related info. Thanks in advance.

Hi, Jade, can you share some of the experiences sometime to us all if you do not mind?

Sure. My first wonder trip consists of discussions with 9 people, appro 1 hour each. I was asked about 60-80 questions on math, finance, econometrics, C++, and a few brainteasers. To name a few:1. To derive closed form solution for perpetual knock up and in barrier.2. MLE stuff.3. local vol and related P&L4. 12 multi-choice on C++, a progamm on vector template & fstream, a programm on multivariate normal simulation5. Derive solution to CEV6. American on American compound7. Credit risk, structural & reduced form models

JP,They do treat you different Fresh PhDs would get easier questions...Any interesting brainteasers you remember from the interview?Hope you'll get your job soon,Olya.

There are just a few brainteasers, as far as I could remember, including:1. fly between two buses2. 3 boxes, apple, orange, mixed, all labled wrongly, how many tries to identify boxes.3. families wanting boys keep on having kids, what is the distribution and what is the eventual expectation.Anyone here has experience in model validation? Any interview hint?

QuoteOriginally posted by: JadePoissonAnyone here has experience in model validation? Any interview hint?It will depend on whether it is for FI, equity, credit, or something else. I can give you some hints if it's for FI.

Thanks! I would love to hear.

QuoteOriginally posted by: JadePoissonThere are just a few brainteasers, as far as I could remember, including:fly between two busesI assume this is the well-known problem of a fly zigzagging back and forth between two objects moving closer together. It's easy to write down a geometric progression for each distance and sum it, but you can just get to the answer directly from the relative speed of approach and fly's speed.Someone once asked John von Neumann this puzzle and he instantly gave the correct answer. "Yes." said the questioner, a little disappointed, "Most people fall into the trap of summing the infinite series." von Neumann (slightly puzzled): "That's how I did it."

Yes it is. I heard about Neumann's story before, actually I heard many his stories. Quote Ulam: "John is a person who knows how to live to his fullest, but don't know how to die." Neumann is so smart that I often feel bad when I read his papers or stories.

If it is for the FI model validation then you might be ask anything about 1. products (swaps, caps/floors, swaptions, and if you know those, may be something more sophisticated: CMS, TARN, PRDS, Snowballs, Range accruals - the list is endless) 2. models (one-, multi-factor short rate models, HJM, BGM, SV, etc.), their advantages and drawbacks, calibration issues, dealing with optionality, early exercise, path-dependence. 3. tests (whatever you actually going to do to test a new model)Of course, this is in addition to all general questions (math/stat/finance/programming/brainteasers).Hope this helps,Olya.

Olya, Thanks a lot. Your post is definitely helpful. JP

Btw, yesterday I happened to come across two most impressive interview questions.One is a brainteaser: Can You price this building by BS formula? If yes, how? If not, why?The other is quite advanced math: given two optional processes X and Y, for any stopping time tau, E[X(tau)]=E[Y(tau)], are X and Y indistinguishable?

On the second question, one is tempted to reply no. But if you define a stopping time tau_{epsilon} := inf{ t\ge 0; sup_{u in [0,t]}|X(u) - Y(u)| > epsilon } (should be a stopping time as X, Y adapted),then you could be in business when you consider the intersection of the set of stopping times {tau_{2^-k}}.

Etuka, your answer is cute. But the question might be a bit harder. In your solution, you did not use the key assumption: X and Y are optional. The classical solution is to use the section theorem. At any rate, I am not sure how such a result can be used in Finance but the interviewer told me that he actually used it couple of months ago. I was very impressed. JP

An optional process is adapted and regular, right? The adapted bit is used in the stopping time def. and the regularity is presumably used in the handwavy convergence bit. I have no doubt you are right about the question being harder - I handwaved precisely because the devil is in the detail...I can't believe you've made me look up the section theorem...