2 Days in the Life of a Risk Validator
Вставка
- Опубліковано 26 тра 2018
- Risk management is a mystery to most people. A lot of people assume you use VaR for everything but the truth is you don't use it often if you are not on the investment side. This video is two days with me as a risk manager in a large bank. I work in the office 3-5 days a week but for the purpose of this video I am working from home (whf).
Takeaways:
1. Risk management is mainly statistics focused.
2. Technology is usually the largest hurdle for productivity.
3. There is a lot of collaboration between colleagues and departments to make an effective risk management framework.
That's really amazing... Doing WFH way before covid... I agree with you on many advantages it has but it is so difficult to convince many others about them....
Just one word - you are "AMAZING"- Managing your work and sharing your knowledge for the world!! Superb!
Thanks! I couldn't do it without views/subscribers like you.
Just found this channel. Really interesting. I see that you discuss analytics and machine learning while working at a bank. I have been a quant supporting the exotics derivatives desk at a bank and have never used that. In fact, I am sure my coworkers and I don't know anything about ML but the word. What type of models do you work on which uses ML at a bank? If you validate derivatives pricing models, I don't think you would use that at all.
GBM has been picking up in popularity for credit modeling. Other uses are for business operations like NLP.
Hey Dimitri , Great video .Thanks .how do you decide what model to select while valuing a client if its not a VaR based measure.also this might be a broader question , my understanding are these measures are used to fundamentally implement a capital charge on the desk in the form of an XVA perhaps , therefore given this , do you value the client different ways and then take the least or highest charge and then charge this to the desk ,because that would potentially also impact the pnl from a trading perspective right , given how much a desk might be constrained dealing with a client .Thanks .
For derivative products you would use XVA as you have mentioned. The XVA approach still prices derivatives with more standard methods such as the Black-Scholes but then adds the adjustments. In banking (making loans) there is not a standard method or formula as there is in pricing derivatives. For pricing loans developers look at different personal characteristics that can predict the different components of expected loss (EL = LGD * PD *EAD). The models to price those components are statistical models such as logistic. Examples of variables could be debt to income ratio, FICO score, and other bureau data.
Thanks Dimitri.
Interesting video! Thanks for making this.
I love your videos, Dimitri. Such an approachable channel. I am a software developer going back to school to study applied math. I'm hoping to break into an quantitative research or development internship. Your channel is very motivating. I am trying to hone in on probability, stochastic processes, and mathematical finance courses for my electives. Would you say that I'm on the right track?
Topic wise, you are on the right track. Remember a masters degree is a minimum, at least in the US. Quantitative finance is a lot of statistics.
Thanks for the feedback as well!
I just came across this video and others, i must say that they are so helpful.
@Dimitri could you please advice me on this. i am doing my final year in chemical engineering. but have applied and been accepted to do BCom Economics and Statistics majoring in statistics, i would love to be a quant one day after also doing the pure math core modules. which will merge as a BSc Math n Stats degree. so would you say am on the right path or do i need to do more than what i am doing. please advice.
Your path is a possible path to quant finance. In your grad program make sure to take as many statistics classes as possible. Risk management and much of quant finance is just statistics.
Working from home is SWEET
Hey Dmitri, awesome video by the way! May I know what kind of model (math wise like machine learning or maybe a theoretical finance model) does your team validate?
We validate anything that is submitted to us. As some examples;
logistic, fractional logistic, OLS, pooled OLS, SARIMAX. fixed effects, random effects, WGLS, DtD, CART, random forest, and GBM trees.
@@DimitriBianco Oh okay, got it. May I know who does this model development work? And is there interesting work to be done there or are there a lot of constraints because of regulations faced by the bank?
These jobs are actually called, "model development." The work is interesting and regulations can make the process more interesting and challenging. If you search online at a big bank for jobs called model development you'll find what they look for in hiring.
Hey Dimitri! I am machine learning developer since last two years and now I am switching my career to finance. I have just completed Quantopian lecture series and building algos for fun on Quantopian and QuantConnect.
Any tips on how should I get a job with a very basic knowledge on quantitative side?
Also, I am considering joining WorldQuant MScFE program part time. What are you views on that?
Thanks a lot. Your videos are very helpful!
My advice for getting a job in the quant world with a machine learning background is to look for jobs that already cater to your skills. If you are good at programming then look for a programming job at a quant finance firm. If you are good at statistics then look for a job doing statistics (risk management, stats arb, etc.) in a bank or quant finance firm.
As for the WorldQuant MScFE...it's possible to learn financial engineering online but I think you'll find it very difficult to find a job with an online degree. Most people I work with or have met in my industry have masters and PhDs from the top universities in the US. Most of them are from another country but they end up coming here for an education. There are a lot of financial engineering and other quant programs in the US which generate a lot of graduates. I'm not saying all of these students are good quality but it does make the hiring process hard and most banks will review your education as a filter.
Good tips. Thank you!
Just curious, do you use SAS instead of R since it's the standard for your line of work?
Yes, we use SAS because it's the industry standard. It also provides a lot more useful output with less coding.
@@DimitriBianco Are you currently based out of Canada ?
Hi Dimitri! I am an undergraduate student in Finance and want to work in Risk Management. It is advised to take a Master degree in Statistics or Quantitative Finacne. However, the thing is most of those masters prefer applicants who have a quantitative degree like Maths, Actuarial Science, Computer Science.
Any advice on which master degree should I take to become a Risk Analyst? Thank you very much!
Statistics would be the best route. You can also consider financial engineering (FE) type of programs as well however a top FE program would be more challenging than statistics. If you are worried about program acceptance, applied economics could be a good route however you really have to make sure they have as much math and statistics as possible. Economics isn't viewed as good as statistics however it is a viable route (I took this route).
Thank you so much, Dimitr. All of your videos are so great and interesting, really inspiring me a lot.
Hey! I want to ask you an opinion. What so you think of getting a bachelor in mechanical engeneering and the a master in quantitative finance for example at ETH Zurich?
Doing engineering as an undergrad and then a masters in quantitative finance is fairly common. Engineering will give you the applied math background needed for the masters.
Thanks for you answer! Do you know if the master cours of quantitative finance at ETH Zurich is good compared to the offer in USA?
Hi Dimitri, thank you for making this great video, it gives a concrete idea of what a risk validator does! Apart from SAS, what else do you use mostly? Do you also use Matlab, Python?? Thank you.
We occasionally use R and Python but more for learning than working. One of the most enjoyable parts of my job is doing math by hand to solve issues within data and models. I would have never said that in high school, college, or grad school but I really enjoy it at work and in personal studies.
Hey, I am a student from the UK completing my masters would love a 30min meeting for some advice.
You're a Joy Division fan, very cool.
Sony MDR 7506 headphones. you must do some type of music production in your spare time,no?
Yes, I was a DJ when I was younger. It looks like you are involved with music as well.
Sounded like a cool project! How was the cluster analysis used?
We typically use cluster analysis for variable reduction and variable selection. We start with 1,000+ variables so cluster analysis makes this a lot easier. As a validator I usually issues findings around how many clusters are selected. For some reason a lot of developers don't review the marginal cluster contribution.
Dimitri Bianco Wow that is neat! Thanks for the insight Dmitri!
Hey Dimitri, so which are the criterions that you use to decide the right number of clusters?
No load balancers on their end for scale?
We do have load balancers internally however when the holidays come there are a lot of developers and validators running large programs in a mad dash to finish project before they go on vacation. Most banks (including the one in this video) have separate servers for different user groups as well but once in a while things crash.
Dimitri Bianco maybe it's a limitation of the database itself??
It could be. We have tech people that solve these issues when they arise.
I'd like to know the same, how you're mostly like to use the cluster analysis?
Funny to see you doing the same movement many times when the camera is high speed mode!
A big company and a subsidiary. Do you get paid twice ?
I wish lol
You worked for 2 days from home, but how often are you on site at your company?
Sergey it depends on the type of projects I'm working on. If I have important meetings I'm in the office but if I have a lot of analytical work then I work from home. I try to work from home 2-3 days a week. Working from home is a fairly common perk in risk management.
Sounds good! I have a Data science friend who landed a remote job in a company on the other side of the world. To me this sounds really weird, since I think there is a lot of value of having people sitting around a table instead of skype. Do you have any opinions about such serious jobs being done remote?
For client facing roles I think it is important to be in person or to travel to build in person relationships. When you work in a highly technical area it can be difficult to explain concepts without whiteboards but there are software that can help. On the flip side I love working from home and usually have it as a requirement when negotiating a new job. It's fun to be around people in the office but those fun conversations can waste a lot of time. For someone in risk management I see no need to actually be in the office as long as employees are getting work done. The short answer, yes I think you can serious work done at home but it does depend on the type of job.
Greatly appreciate the elaborate answer.
Hello Dimitri. Nice videos. But your coffee making is inefficient. Use instant coffee instead.
Lol I actually drink instant coffee but only when I'm in a hurry. Nothing beats fresh ground beans and a Moka Pot.
woah, is it the norm to be able to work 1/2 days a week from home? that's pretty cool
We still end up working 8+ hours per day but you are able to manage your time better. Not all banks or companies allow it but a lot of them do. I actually have friend who works from home full time which is somewhat rare.
@@DimitriBianco I guess that is hard to have in big banks .
Hmm Xmas tree was still up in May? XD
Fyi after coffee cools down for 20mins bacteria starts to grow
Lol, this video was shot right after the holidays in January but it's taken me a while to get it edited and uploaded.
Dimitri Bianco heh not judging mate! Love the sweater btw. Cheers
Just wait until you meet the colleagues who never wash their coffee cup, hahahah.
Funny you should say that today...I only just washed my morning coffee cup at like 8pm. BUT I drank it within 5mins :)
Joy division
You cant be serious working with 15 inch HP laptop , and wondering why servers are down.
Lol my colleagues give me a lot of flak for it as well. I'm actually upgrading my home office over the next few weeks and with that will come a larger monitor. At the actual office I have a docking station and two larger screens.
Popular demand...