Navigating Career Change within Data: Q&A with Juhi Singh | Coursera Community
Coursera Header
Q&A

Navigating Career Change within Data: Q&A with Juhi Singh

  • 16 April 2019
  • 26 replies
  • 2452 views
Navigating Career Change within Data: Q&A with Juhi Singh
Userlevel 7
  • Community Manager
  • 1111 replies
Juhi Singh, Data Scientist at Coursera, is taking your questions about working in the Data Engineering domain in a rapidly growing company with different businesses, challenges in architecting scalable data models, etc.

This thread is now closed to new questions.

About the Q&A Host
Juhi is a Data Scientist at Coursera and formerly worked as a Data Engineer for 2.5 years. As a Data Engineer, she worked to democratize data among internal and external stakeholders in service of Coursera's mission of transforming lives through learning. A 3-year Courserian, Juhi has built partner-facing data offerings ranging from the new Course, Specialization, and Admin Dashboards to DataHub. She holds a master's degree in finance from MIT, a master's degree in economics, and a bachelor's degree in engineering from the Birla Institute of Technology and Science. In her spare time, Juhi enjoys hiking and exploring new trails in California, reading, and dancing.

This topic has been closed for comments

26 replies

Badge
Rapid change and intensified competence is prevalence in data science domain. Learning ancient philosophy of your culture help you building core values. It guide your direction when you are confusing.
do you have any advice for individuals currently working in oil & natural gas or an industry that isn't always a front runner in new technology/methods? data science will have a massive impact on our industry, but we're usually slow to adapt to & adopt new techniques & utilize new technologies or platforms. i don't necessarily plan to be come a full blown data scientist (personally), but would like to leverage my existing knowledge of the industry with the pending influx of data scientists & data heavy projects. thanks!
Userlevel 1
Badge
I'm curious to know if there are students data science courses in Coursera who are above 50 years old ? Is the digital tech community ready to accept and work with mid-lifers willing and able to re-skill themselves ? With wealth from the repository of their experience, what do you think we can draw from them ? thank you.
Badge
As you have been both on the data engineering and science sides, I have questions on both domains.
1. Data science: other than clustering algorithms do you see unsupervised and semi supervised techniques widely adopted by practioners from industry who either can't afford to label the data or don't have very large data to leverage on deep learning techniques?

2. Data Engineering : how often do you refractor the pipeline building code to meet the 99.99 uptime requirement? Or is it managed purely by tweaking the load balancers? How do you forecast your public cloud bills in times when the load is uncertain.
Badge
Hi Juhi - thank you for giving all us inquisitive people this opportunity to ask questions. I really appreciate your time.

i see you have a background in finance. I too have an education in financial engineering and currently work in the sell side quant industry. Could you please provide some insight into the path you recommend individuals like me can follow to make a successful transition from finance to data science? In particular, what technical skills do you see being used in the data science industry that are not necessarily front and center in finance? Thank you.
Thanks for having you...
My question is how is it possible to combine Data Science and Cyber Security.
Could you please advice on the pathways available ?
Userlevel 1
Badge
I'm a fairly new data analyst. I work in a field - the utility industry - where data science is looked at as holding a lot of potential to manage ongoing issues (reliability, pricing, etc.). My biggest concern is that I am often asked to make calculations based on assumptions that are not quite correct or a misunderstanding of what that calculation actually means. My question is, what are the biggest pitfalls you see where data science can be or has been misapplied to a problem? Is that a tactic you suggest to communicate better what a formula represents to people who are more verbal vs. numeric in their thinking? Thank you!
Badge
I'd like to build a portfolio of applications that make use of data science. My background consists of machine learning, web development and cyber security.

How valuable will a such a portfolio be when I apply for jobs in data science, how to maximise its benefits in my job hunt 2 years from now, and what aspects of my career search and personal time management do I need to pay attention to while creating these apps? Thanks.
Badge
Hello Juhi, thank you for sharing your experience with us.
My questions are very simple and open:
would recommend a career in Data Analysis to beginners like us ?
I just took a course on Coursera about the basic of Data Analysis
What skills you used to grow in this field ? So what challenges you had to face ?
Thanks very much
Michelangelo
Badge
thanks for helping us on Data Analysis
What background do you expect in a candidate profile ?
I mean I have a degree in Business Economics and then a course in data Analysis, would be a good start?
do you think creativity and passion are good elements in this field ?

(Moved to active Q&A event with Juhi Singh by @Laura)
Userlevel 2
Badge +1
Hi there, Juhi!

In your personal experience.... Any advice you could give us in regarding: algorithmic trading and speed trading would be more than fine, involving stock exchanges and mercantile exchanges...
Badge
Hi!

Any advice for people with a finance background looking to get into data science?
  • Are there a minimum set of skills that should be obtained/demonstrated in order to land a role?
  • Are there a subset of specific roles that one should target in order to land on the data science beachhead more easily?
  • How is the perception of self-learners vs people with degrees in data science?
  • With the quickly morphing landscape in data science, is it a domain that will offer stable employment for years to come or is it more vulnerable to automation than others?
  • What are the salary expectations for someone looking to switch careers into data science (i.e. mid -career - approx 40 years old)?
Thanks!
Userlevel 4
Badge +3
Anybody ever progress into this career without a technical background or a degree? I’m very interested in becoming a data scientist so am self teaching math and statistics while also working through the Kaggle learn material before I start to do real projects. Ideally I’d like to make the transition into this career within 1-2 years. What is likelihood of being taken seriously?
Badge
I have just started learning Machine Learning and I want to be able to got professional in the shortest time possible. What are your suggestions for learning paths and methods that will help me achieve that. Currently I am working in the finance industry as a full stack developer. How viable is for me to change my current work and become a data scientist without a masters degree or a phd.
Badge
Hello Juhi Singh and Laura. Thank you for this post. 🙂 It will be very useful.

My questions are regarding one specific role in the Data Scientist area: Data Architect

I am at the Data Science world since 2 year ago. I started has an ETL trainee.
Meanwhile my boss told that with my personal and professional characteristics I will be an excellent Data Architect - A role that I never heard before he mention.

Then I saw this fantastic article from DataCamp:
https://www.datacamp.com/community/tutorials/data-science-industry-infographic

My questions are these:

  • In your opinion what should be the Data Architect characteristics? - either in terms of technical knowledge or personal characteristics.
  • Which should be a good study path to be a Data Architect?
There are lots e-learning courses that focus on Data Analyst and Data Scientist career path (for example in DataCamp and Coursera) but it looks like that to be a Data Architect we must take individual courses in several MOOC/schools so we get knowledge about ETL with python, Data Warehouse, Big Data courses, etc.

  • And for last, regarding the previous question do you recommend any good online courses that I should take?
I know these weren't simple questions but the information about Data Architect is so confuse and different in several articles/blogs/forum that I am totally confuse with this role.

Thank you very much in advance :)
Best Regards
Badge
Hii,
Thanks for helping us.
I have done done bachelors of business administration from India and would be doing my Masters in Management Information Systems this fall in US. I also have no work experience. I am currently learning Python and SQL from coursera, and I am planning to do the IBM certified data science specialization. My questions are as follows:
1) I understand that for a fresher a role like Data Analyst/Business Analyst is more accessible. Are their certain technologies/skills you would suggest someone like me to learn so that I can get an internship/full time job in data science domain.
2)Since you have done your masters from US, are their certain tips you would like to suggest to incoming grad students like me, in order to successfully land a job?
Userlevel 7
This thread is now closed to new questions. @juhi.singh will be replying to your questions shortly!
Badge
Rapid change and intensified competence is prevalence in data science domain. Learning ancient philosophy of your culture help you building core values. It guide your direction when you are confusing.

Hi freetest2019,

Thanks for that note - I agree with your sentiments. Irrespective of your function in your organization, I think it is imperative that we consciously decide what our core values are and how are we letting it shape our worth ethics. As data scientists, we provide our teams with the insights that shapes decisions and the company's growth. Topics like privacy concerns and algorithmic fairness are important to be kept in mind. If you are interested, I would recommend the Data Science Ethics course as a good place to understand and think about this topic.

Thanks!
Juhi
Badge
I'm curious to know if there are students data science courses in Coursera who are above 50 years old ? Is the digital tech community ready to accept and work with mid-lifers willing and able to re-skill themselves ? With wealth from the repository of their experience, what do you think we can draw from them ? thank you.



That’s a great question! I personally believe that implementing data science skills doesn’t have to involve new technologies, you can always leverage your current platform and derive more insights from your data by equipping yourself with new data science skills. Given that you already have industry experience, here are a few things that I would recommend:

  • Getting comfortable with data: Building a solid foundation of database manipulation skills, SQL, Python/R and statistics goes a long way in getting started with data.
  • Get some real world experience by applying these skills to datasets that you have are already familiar with. As you get better with using these skills, it becomes easier to start thinking about the various ways in which you can leverage your data
  • If your company allows open source tools, try to incorporate them. They are fun to learn and a good way to keep pace with what is happening in the data science world!
Hope that helps!
Juhi
Badge
I'm curious to know if there are students data science courses in Coursera who are above 50 years old ? Is the digital tech community ready to accept and work with mid-lifers willing and able to re-skill themselves ? With wealth from the repository of their experience, what do you think we can draw from them ? thank you.

Hi fwlim,

Thanks for that question!

There are thousands of unique learners at Coursera who have self-reported their ages to be over 50 and who have enrolled in data science courses, and we see them completing courses with the same enthusiasm as learners or other age groups :)
I can’t speak for the entirety of digital tech community, but at coursera, we believe that anyone anywhere can transform their lives - in this context, by re-skilling themselves to make them ready for the future. There are several articles that talk about how AI will replace 40% of the jobs in the next 15 years or so. This is where flexible learning platforms such as Coursera shine best.

I think if you have strong problem-solving skills, can bring domain expertise to the table and can demonstrate your skills and interest in the field. You could start off with analytics - data visualization, business intelligence to begin with, and then move on to more technical domains within data science.

Thanks,
Juhi
Badge
As you have been both on the data engineering and science sides, I have questions on both domains.
1. Data science: other than clustering algorithms do you see unsupervised and semi supervised techniques widely adopted by practioners from industry who either can't afford to label the data or don't have very large data to leverage on deep learning techniques?

2. Data Engineering : how often do you refractor the pipeline building code to meet the 99.99 uptime requirement? Or is it managed purely by tweaking the load balancers? How do you forecast your public cloud bills in times when the load is uncertain.


Hi xavyjs,
That's a great question!

  1. IMO, semi-supervised learning has a lot of potential, since cost of labeling data can get expensive with an expansive dataset. Rather than labeling all of the data, it is easier to label a small set of the data, train a model on that, predict the output and train. I can't speak for industry standards as I myself am pretty new to this field 🙂
  2. We have a regular on-call schedule to make sure that our uptime/turnaround time meets our SLA. We typically make sure that updates from pipelines are transactional or can be rolled back, when this data serves online services such as recommendations. The online stack is a microservice architecture, and as those machines are stateless, we auto-scale and load balance quite a bit there, as our load varies throughout the day. Keeping our databases over-provisioned for this purpose helps.
Hope that helps!
Juhi
Badge
Hi Juhi - thank you for giving all us inquisitive people this opportunity to ask questions. I really appreciate your time.

i see you have a background in finance. I too have an education in financial engineering and currently work in the sell side quant industry. Could you please provide some insight into the path you recommend individuals like me can follow to make a successful transition from finance to data science? In particular, what technical skills do you see being used in the data science industry that are not necessarily front and center in finance? Thank you.

Hi rajat,

I can relate to this question a lot since I also have a similar background. Most financial engineering programs cover atleast one programming language, along with basic statistics skills like regression modeling. Personally I have felt that the technical skills taught in financial engineering schools are enough to get through the door - you may have to pad it up with data science projects to show your interest.
When I was gearing towards data science after my finance degree, this is what helped me fill gaps:

  1. Start with SQL, if you worked with databases before. SQL is easy to understand, and is used by almost every data scientist retrieving data from tables and manipulating them. I really like the SQL for Data Science course on Coursera.
  2. Python: I used Python in my finance degree a lot, but as I did not have a CS background prior to that, it was helpful to get a refresher in Python.
  3. Statistics: I'm sure you must have covered stochastic calculus, distributions, estimations and non-parametric techniques, but it is also helpful to understand basic statistics like hypothesis testing, regression modeling, etc.
While there is quite a bit of an overlap between financial engineering and data science, some topics like Black Scholes is very specific to financial engineering, whereas other topics like hypothesis testing and experimentation is more prominent in data science.

Thank you,
Juhi
Badge
Thanks for having you...
My question is how is it possible to combine Data Science and Cyber Security.
Could you please advice on the pathways available ?


Hi nash_tgc,

Thank you for the question! In my view, data science is a pretty broad field, and there is already a lot of data science applications that are used in cyber security automation.
From my limited knowledge of cyber security automation is a combination of threat monitoring, followed by accurate detection and timely response. One can use Data science to improve on all three fronts by making smarter threat monitoring, detection and response systems. Although I'm not an expert at cyber security, I would recommend this post as an interesting read.

Thanks!
Juhi
Badge
I'm a fairly new data analyst. I work in a field - the utility industry - where data science is looked at as holding a lot of potential to manage ongoing issues (reliability, pricing, etc.). My biggest concern is that I am often asked to make calculations based on assumptions that are not quite correct or a misunderstanding of what that calculation actually means. My question is, what are the biggest pitfalls you see where data science can be or has been misapplied to a problem? Is that a tactic you suggest to communicate better what a formula represents to people who are more verbal vs. numeric in their thinking? Thank you!
Hi A. Jenner,

That's a great questions. I agree with you, there are times when people want to justify an answer with data, rather than look for an answer within the data.
Some of the biggest pitfalls I see is:
1) Having an ambiguous hypothesis, and not explicitly mentioning what the assumptions are.
2) Starting off with a biased dataset, and training your models on it. This sometimes leads to stereotypes being perpetuated in your analysis and recommendations
3) Not being able to explain the nuances and outcome of your model/analysis at an intuitive level, and solely relying on the model results.

Most of these pitfalls are mistakes that you should be aware of while producing your results. However, it is equally important that we be honest in our reporting and interpretation of results. There is a famous quote that "If you torture the data long enough, it will confess to anything". I've seen issues like selective reporting (So and so company was no. 1 in ranking *but only for that year, with several other filters applied), not actively working on removing selection bias when working with particular datasets such as surveys, etc.

On the verbal vs. numeric aspect, I'd recommend sticking to simple visualization, leveraging the use of dashboards rather than showing code, and being able to explain in an intuitive, actionable fashion what the data suggests is helpful. Before communicating, ask questions such as:
1) Who is your audience? Put yourself in their shoes and try to understand where they are coming from. Empathy is a great skill for a data scientist to have :)
2) What action are they going to take out of your presentation/communication? Is it a technical outcome (eg. updating recommendation systems), is it a strategic/business one (eg. changing the marketing focus) or logistics/operations based.
3) Do you want to save all of your outcome/analysis for one final presentation, or provide it in an piecemeal, iterative way? I would lean on the latter

Once you have these jotted down, it would be much easier for you to be able to communicate the right results in a way that is convincing and easy to grasp.
Hope that helps!
Juhi
Badge
I'd like to build a portfolio of applications that make use of data science. My background consists of machine learning, web development and cyber security.

How valuable will a such a portfolio be when I apply for jobs in data science, how to maximise its benefits in my job hunt 2 years from now, and what aspects of my career search and personal time management do I need to pay attention to while creating these apps? Thanks.


Hi yieng,

That sounds like a great plan! Having a combination of machine learning, web development and cyber security in your applications would give you an edge over the standard portfolios that people have. One advice I'd like to give here is that instead of going very deep into one project, it would be helpful to showcase your data science and machine learning skills in a variety of medium sized projects. Building an application end-to-end is time consuming. If you want to do it as a hobby, that's great, but if you want to be able to leverage it for your job hunt, and given the time constraints, I'd recommend to scope your projects well before investing in them. One advantage of working on Kaggle projects are that they give you clean datasets and take out the hassle of looking for a place to start. By only focussing onthe data science aspects of the projects, you could invest the time you save in picking up new skills. Most entry level data scientists have a broad base of general knowledge like programming and statistics, regression and ML models and specialize once they start the job. I'd highly recommend to get to that stage first, and if time permits, focus on building apps.

Thanks,
Juhi