Preparing for the job interviews is a tedious process. It can be more difficult for data scientist positions than other tech positions, since the interview could cover a wide range of contents, including but not limited to statistics, coding, product questions, behavioral questions, etc. But if you ask me what you would be asked the most during the interviews, my answer is: “talk about one DS project that you did”.

Well, I think working on data science projects is one thing, and knowing how to talk about them is another. In this blog, I hope to share with you some techniques of how to effectively talk about your DS projects in an interview, and how to prepare for the talking.

1. It is a communication, not a chat!

What is the difference between communication and a chat? In communication, you are talking with a purpose to your audience to convey certain information that you want them to know. While in a chat, it’s more like unstructured talking without a clear purpose, and you could allow your topics to jump here and there. Therefore, you do need well prepared for communication but may not need to do so for a chat.

So what information that you want to convey by talking about your DS projects?

The goal is simple and clear: you want to show the interviewers that you are capable to do the job well. To achieve this goal, you need to show the following skills: 1) Technical skills: Python, R, SQL, Spark, AWS, etc. 2) Knowledge of data science: machine learning, statistics, ETL, data visualization, etc. 3) Problem-solving skills: how you dealt with difficulties working with the real-world data that were not expected at the beginning of the project 4) Communication skills: how you worked with others including team members, stakeholders, project managers, etc.

It sounds challenging to include all the information in 5-10 minutes to talk about your projects. But no worries, I will illustrate below how to include everything in your talking.

2. Know your audience!

Yes, it does make a difference when you are talking about your projects to interviewers with different roles. Usually, you would interview with four types of people: HR, team manager, data scientists, business partners/stakeholders. The total interview time window varies depends on who your interviewers are. For example, an HR call usually takes 15-20 minutes while an interview with a data scientist probably takes 30-60 minutes. Therefore, you need to pay attention to the time you spend talking about your projects depends on who you are talking to. The following table demonstrates the general guidelines of how to talk about your projects with different interviewers.

As you can see from the table, you need to prepare for different versions of answers to the question of “tell me about your DS projects”. You need to have good control of the flow while talking, emphasize different areas and not give too much or too little technical details. It is challenging!

So next I am introducing a strategy/method I used during the interviews, with section 3.1 talking about the 4-step method I use, and I will show you an example in section 3.2. Let’s get started!

3. How to talk about your DS projects?

3.1 The 4-step method

You probably have known or heard about the “STAR” method to answer behavioral questions in interviews. I found it can be helpful to talk about your project, with some tweaks of the method. Based on my experiences, I summarized the four steps to follow when you are talking about your projects:

Step 1. Project background and objectives

This is how you start talking about your project: by providing some background information and point out the goals or objectives of the project. This is like the “Situation” and “Task” in “STAR” method. The goal of Step 1 is to give your interviewer basic knowledge about your project, but you do not need to spend too much time on this part. Depends on who you are talking to, the background information can be one sentence or a few with some elaborations. And you will only need one sentence to describe your main goal.

What to talk about for “Background”: your role in this project, did you work with any team members and/or business partners for collaboration, necessary explanation of business needs and/or why the project is valuable. You could also mention a bit of the coding language or tools you used.

What to talk about for “objectives”: you may have multiple objectives for big projects, but I suggest you state no more than two main objectives. Limited by the interview time, so you should make this part succinct and strong for interviewers to follow easier. You could state your objectives in one sentence with the format like: “the goal of my project was to develop an ML model/DS product/create a tool/evaluate the impact of XX”.

Step 2. Challenges and solutions

I would say this is the most important part of your story and also it is the part you would get most of the questions from the interviewers, so be fully prepared and practice! You need to talk about “Action” as in the “STAR” method, but you need to tell a good story here. That’s why I found talking about challenges and solutions is better than just listing what work you have done. When you mention the “challenges”, it can really attract the interviewers’ attention since they would want to know 1) how you define challenges, which represent your skills and capabilities, and 2) your problem-solving skills like how you handle difficulties in work.

What to talk about for “Challenges”: I would recommend talking about one technical challenge and one business-driven challenge. If the project is your side project or a course project without a business scenario then you could focus on the technical challenges only.

Examples of technical challenges include dirty data, insufficient data, weak correlations between predictors and response variable, imbalance dataset, modeling difficulties, deployment difficulties, etc.

Examples of business challenges include tight deadlines, short turnaround, difficulties to communicate with non-technical business partners, change of project scope, etc.

What to talk about for “Solutions”: after talking about the challenges you had, you could start talking about the solutions with an introducing sentence like “To deal with these challenges, I did the following”. Then you could illustrate what you have done to deal with the challenges you mentioned previously. Remember to use words like “First, second, then, last” to make your points sound and organized.

Step 3. Achievements

Achievements are a little different from the “Results” in “STAR” method. It’s not just listing the results objectively, it is the part for you to “show off” what you have achieved because it’s very common for you to get the question of “why you are proud of this project”. Again, you could tackle this part from perspectives of both technical and business-oriented. List two or three achievements would be good enough since you don’t want to spend more time talking about achievements than your work (in Step 2). Make sure that your statements are strong to impress the interviewers.

What to talk about for “Achievements”: Talk about the impacts, would be better if you have numbers to support your point. From a technical perspective, you could compare your model to previous models/methods in terms of increased accuracy, running time, etc. If the model has been put into production, you could mention numbers of model users, positive feedback from the users, etc. If it is a personal project on your Github, you could talk about the number of stars you got. From a business perspective, sales lift, revenue, loss prevented are all good numbers to talk about.

Step 4. Lessons learned

This step may be skipped if you are talking with HR or have a tight interview schedule. But I suggest you prepare for this section to show the interviewers that you are able to reflect and learn from your experiences. Again, one or two strong statements here are good enough.

What to talk about for “Lessons learned”: Talk about skills you gained, but need to be relevant to your story in Step 2. For example, skills like developing ML models, making advanced plots, creating ETL process, etc. The skills do not have to be all technical, and you can add soft skills like negotiating with the clients, communicating with non-technical business people, etc.

3.2 Example

After all the theories of the 4-step method mentioned above, I am pretty sure that you have got a general idea of the process. Now I hope to illustrate with an example below*. Please note that this example only covers important points and more details need to be added in the real interviews.

*(Step 1 Project background and objectives) I’d like to talk about my project of identifying fraud transactions. First, I hope to provide with you some background information. I am a lead Data Scientist for this project and I work with business partners from the sales department. The goal of the project is to use historical data to develop an ML model to predict if a transaction is a fraud.

(Step 2 Challenges and solutions) I had two challenges with this project. The first challenge is that the historical data is extremely imbalanced, because we only had 1% fraud among all the transactions. The other challenge is that my business partners do not have technical backgrounds, so I need to learn the best way to communicate with them to convey my findings. To deal with these challenges, I tried a few different things. For the imbalanced dataset, I used SMOTE sampling and assigned weights in ML models to handle the imbalance. I also had a Coursera class to learn how to communicate more effectively within a business environment.

(Step 3 Achievements) I am really proud of this project because the ML has saved $XX since it’s been put into production. I also got positive feedback from my business partners that the model is easy to use with an overall accuracy of XX%.

(Step 4 Lessons learned) By working on this project, I have practices my skills of developing classification models with imbalanced data, and I have accumulated experiences in presenting results and analysis to non-technical people.*

4. How to prepare for the interview?

Tip 1: don’t wait until you are ready to find a job. Prepare early!

This is the most important tip that I hope to share with you. The reason is simple: we forget things, so it’s better to write things down when they are still fresh in your head. I learned this from my own experiences, during one interview I was asked about details of a Design of Experiment (DOE) from a project I did over 5 years ago. Though I could tell the interviewers it’s been too long so I could not remember the details, why would I let this happen in the first place? So to avoid this happens again in the future, what I do is when a project is finished, I would document the necessary details in OneNote. Of course, you could use Word or Google Docs or whatever you like, as long as you record the useful things that can help you in future interviews. Please see the pic below as a checklist of documenting your work, which includes the important points mentioned above in section 3:

Please note that do not document sensitive information that is forbidden to disclose by your company, and do not talk about that either in the interviews!

Tip 2: Practice and time yourself.

  • First, you could start practice talking by looking at the notes you documented for the projects. I don’t recommend writing down every sentence you want to say, but you could add key sentences and keywords to your notes. This ensures that you have an organized way to talk.
  • Practice and prepare for different types of interviewers, keep in mind the differences mentioned in section 2 of this blog, so you would know which part to emphasize and which to skip depends on the interviewers.
  • Once you nail down the contents to talk about for different audiences, record the time when you practice and make sure that the length of your talking is within a reasonable time range (refer to section 2 of this blog)
  • Have your friend or family act as your interviewers and ask for feedback about your talking speed, content, structure, tones.

Tip 3: Be familiar with every technical detail you mentioned.

Data Scientist is a technical position, so you must first impress the interviewers with technical skills. Therefore, you need to make sure you are crystal clear about everything you talk about, especially the technical details. Be prepared for questions about the theory of an ML model, or the definition of the model metrics you used to evaluate the model performance. Commonly, the interviewers would start asking theoretical questions about your project.

4. Conclusions

Finding a dream job always isn’t easy. I believe every big improvement comes from baby steps that you have taken. I hope this blog could inspire you to communicate better during the interviews and show the great work you have completed.

Disclaimer: I created this example in section 3.2 for illustration purposes only in this blog. It’s not related to any of my projects at work, and all the numbers and situations stated here are made up as an example.