Suitable roles for a PhD graduate (in computational fluid dynamics)

[alovya]

Can I become a buy side quant researcher with the following background: final year PhD skilled in finite element methods, C++/CUDA/Python and algorithm development?

These skills are used for pricing models [thread 1, thread 2], but my impression is that pricing is a decaying field.

Instead, this post says:

"... the vast majority of buy side quants even at options trading firms will not work on the options pricer itself but focus more on market making algorithms, signal analysis, predictions ...".

1. Would a buy side firm be willing to take me on and let me learn on the job?
2. If I do a data science project, would this be evidence of sufficient statistics knowledge?
3. Can I join a buy side firm as a quant developer, learn statistics in my own time and then move to quant research?

Edit 2: Dear mod, I am really sorry about the edits but I think my original post was too dense.

 

[alovya]

-

 

[alovya]

Any advice would be really appreciated, thank you.

 

[CrossGamma]

I think your profile would be considered for graduate quant research roles. Most Ph.D. graduates are not hired for their specific research experience but their demonstrated ability to work relatively independently on complex problems. You bring some relevant experience related to numerics and while indeed not the focus at most buy-side places, it is a positive selling point. Your interviews will likely still focus on probability and statistics and would recommend brushing up on these a bit. I would not apply for quant developer roles initially (as long as you didn't exhaust your quant research options). While there is some mobility, I have not seen it very often. Also, while the role is called "quant" developer, you will not necessarily work on quantitatively interesting topics.

 

[alovya]

Thanks for taking the time to respond. Your reply is encouraging for me, although I have a simple follow up question about one of your points:

Your interviews will likely still focus on probability and statistics and would recommend brushing up on these a bit.

Would you be able to tell me what level of probability and statistics is expected from me at interviews?

For example, would it suffice if I were able to comfortably answer all the questions from Chapter 3 of this document? Or would I have to be able to explain the maths behind machine learning models to a decent level, for instance?

If you could tell me this it would be very helpful: I can begin interview prep right now if I don't need to know the maths behind ML models. If I do need to know the maths, then I have a lot more left to study.

 

[CrossGamma]

If you could tell me this it would be very helpful: I can begin interview prep right now if I don't need to know the maths behind ML models. If I do need to know the maths, then I have a lot more left to study.

This will depend on the role. For general options quant research roles, the questions in the chapter you reference are close to what you typically see in interviews. If you apply for a more specialized role involving machine learning, then you can expect questions on this too.

 

[alovya]

I see. I'm assuming that the specialised roles involving machine learning are the most interesting/well-paid quant research roles at the moment. In that case I will have to take an ML course and do a personal project to pick up some of the necessary skills?

 

[CrossGamma]

I see. I'm assuming that the specialised roles involving machine learning are the most interesting/well-paid quant research roles at the moment. In that case I will have to take an ML course and do a personal project to pick up some of the necessary skills?

I don’t see this being the case, no. The best-paid roles are typically the ones closest to trading. I personally use very little machine learning on a day-to-day basis. Especially in options trading, people care about the interpretability of their signals and position. The edge is often in designing good features and predictors as opposed to using complex models.

 

[alovya]

I don’t see this being the case, no. The best-paid roles are typically the ones closest to trading. I personally use very little machine learning on a day-to-day basis. Especially in options trading, people care about the interpretability of their signals and position. The edge is often in designing good features and predictors as opposed to using complex models.

I understand. Thanks a lot for your advice, you may see me post again on this forum with more questions as I get deeper into the application process. Hope you have an alright weekend )

 

[Daniel Duffy]

It's not every day one comes across someone with FEM and computational fluid dynamics. It was an area I worked in both in academia and industry It turns out that my PhD research in 1980 could be direct;ly applied to PDE in finance. I have written up here for FDM.

Numerical Methods in Computational Finance: A Partial Differential Equation (PDE/FDM) Approach

<p>This book is a detailed and step-by-step introduction to the mathematical foundations of ordinary and partial differential equations, their approximation by the finite difference method and applications to computational finance. The book is structured so that it can be read by beginners...

www.wiley.com

(no FEM in book because no one knows it??)
BTW was your FEM blue sky stuff or working on existiing FEM packages?

With your background, it should be possible to make an impression based on C++ and maths knowledge.
And if your underlying numerical analysis is sound enough then ML and data science (whetever that is..) should be doable.

My 2 cents.

 

[alovya]

It's not every day one comes across someone with FEM and computational fluid dynamics.

To be honest, it's the Discontinuous Galerkin method, which is only "technically" FEM: practically speaking it's FVM but using linear instead of flat shape functions.

BTW was your FEM blue sky stuff or working on existiing FEM packages?

I wrote all my code from scratch and my PhD is about parallelising this model on a GPU.

To make the parallelisation possible, I implemented a new algorithm and a new data structure and made adaptive mesh refinement fully resident on the GPU (no CPU/GPU data transfers), which is the crux of my research.

My research is sadly not about mathematical innovation but I do like maths though I've never taken a proofs-based course.

With your background, it should be possible to make an impression based on C++ and maths knowledge.

I don't know if linking to one's GitHub is allowed on this forum but mods, please let me know if not.

This and this are examples of the C++ code I've written, though unfortunately I can't share the main, more complicated C++/CUDA code I have developed until I'm published.

And if your underlying numerical analysis is sound enough then ML and data science (whetever that is..) should be doable.

I can read and follow a numerical methods paper, but I have no formal training in numerical analysis (or maths in general). Do you think it's worthwhile for me to do some proof-based study of numerical analysis? If so, have you got a learning path? I may start with real analysis.

 

[alovya]

@Daniel Duffy I noticed that you liked my message - does this mean you think that I should do some proof-based study of maths? Or any other comments. Thanks.

 

[Daniel Duffy]

Hi alovya,
I was planning to writing up a small response offline indeed. I am not in IB or a headhunter but my focus is on learning important stuff in maths, C++,. Python. My online courses are on www.datasim.nl and I do much of everything. Have a look at your convenience and let me know if you have any queries. regards D

 

[alovya]

Hi alovya,
I was planning to writing up a small response offline indeed. I am not in IB or a headhunter but my focus is on learning important stuff in maths, C++,. Python. My online courses are on www.datasim.nl and I do much of everything. Have a look at your convenience and let me know if you have any queries. regards D

Great, I will take a look at your courses, which do indeed look like much of everything. If you had any other thoughts e.g. about my research or coding stuff, I would be happy to hear them.

 

[Daniel Duffy]

Thanks. I started maths in 1971 LOL. I got my 1st FEM course in 1974.

I would say major blocks in my Pure maths course (analysis) are vital.

Regarding numerical analysis, it is not necessary to know the theoretical stuff but more a good hands-on feeliing on how to use the methods and after that reading more into the stuff. Then it starts getting closer to ML.

I am just wondering if all these high-potential data scientists understand the underlyng maths.

Online Courses :: Datasim

www.datasim.nl

If you want to dig deeper then this maybe

Online Courses :: Datasim

www.datasim.nl

 

[Daniel Duffy]

Last but not least, like Toyota and Carlsberg , the two Quantnet C++ courses are probably the best on the planet I have been told

 

[alovya]

Hi all,

I'm sorry for digging up this old thread but I thought it would make sense to post on here because the thread already has a lot of background information about me. I also apologise in advance for the long post and for asking so many questions.

Background

I’m a PhD student in computational fluid dynamics at a non-target university in the UK, and I want to apply for full-time quant research (QR) roles (not quant dev roles). I have heard that hiring for these roles starts in August and ends in January (at the latest), i.e., hiring is going on right now. However, I can’t apply to these roles just yet because I assume interviews require stats/machine learning skills, but my skills are limited to numerical methods (finite difference/volume) and C++/GPU programming. My stats skills are at highschool level except for basic stochastic calculus, linear regression (OLS) and standard Monte Carlo (from a previous full-time non-quant finance role). I have no experience in machine learning either, other than Andrew Ng’s 11 week Coursera course in Octave/MATLAB which I did more than three years ago. So, I have four questions:

Questions

1. Is there any hiring for QR roles other than between Aug - Jan?
2. I have gathered that firms in the derivatives space might see some value in my numerical methods/programming skills. Banks (the sell side) are therefore firms for me to apply to. However, what are the names of the buy side firms in the derivatives space?
3. I plan to improve my stats skills by going through and coding up the first four chapters of “Elements of Statistical Learning”, i.e., to thoroughly understand linear regression and classification. Is this plan enough to improve my stats skills to a level to get me past interviews?
4. Here is the link to my CV. Is it enough to get past the CV screening stage and get me interviews?

Closing remarks

Although my questions are about QR roles, at heart I’m interested in any role that has mathematical modelling and direct commercial application, the keyword being “direct”. This is why at the start of my post I said “not quant dev roles” for fear of being put into a dev role that is not directly tied to the commercial side. I am open to a very “close to the money” quant dev role, but I think these are hard to identify. I also feel that if I say I’m open to quant dev roles, recruiters will try to place me in any old quant dev role whether it’s close to the money or not. To finish this up, I would like to ask if you can think of other roles with maths modelling and commercial application. For example, roles in energy trading, commodity trading and operations research also look interesting to me. I would really appreciate if you could tell me the names of any companies with these kinds of roles.

Thank you very much for reading!

 

[Daniel Duffy]

Energy markets might be worth looking into: (spread-style ) options and risk but more complex than in finance.


//
Python
ML not there yet.
Statistics always important!

 

[EngTurnedQuant]

I think your background is enough to get you some interviews but I'd focus strongly on developing some Machine Learning skills (through Coursera or edX). Also, it doesn't seem like you have a background in statistics and probability. You'll need those for interviews in the quant space and for all types of equity and derivative pricing. You can take a look at MIT's Data Science Micromasters program through edX. I completed this and it had four courses, one in stats, on in probability, one in machine learaning and one in data science.

I'm a PhD in mechanical engineering with my background in fluid mechanics (thin film flows) and dynamics. I wrote my own FEM/FD code for a portion of my PhD which solved the coupled PDEs for fluid flow, heat and elasticity in my specific application. Regardless, I've gotten very little traction applying to jobs with trading firms and hedge funds.

Highlight your experience with Python and C++If your heart's set on this, apply to about 50 places and don't be surprised to hear crickets (i.e. no replies) from most. I think I applied to about 30 places and got one interview.

 

[alovya]

It's not every day one comes across someone with FEM and computational fluid dynamics. It was an area I worked in both in academia and industry It turns out that my PhD research in 1980 could be direct;ly applied to PDE in finance. I have written up here for FDM.

Hi Daniel, some of my PhD work about FEM/CFD is finally published: here's the paper and here's the code. Curious to hear your thoughts positive or negative, if any.