Why Data Scientists Need to Prepare for the API Design Interview

The API design interview is a new-ish phenomenon in the tech interview loop. Therefore, you might be wondering if it will apply to you in your data science interview process. In the past - the two roles that most needed to prepare for an API design interview are front-end developers and back-end developers.

That said, almost any professional developer, including data scientists, should study and prep for API work.

To optimize an API, you need to walk through the entire lifecycle of a request, from the front-end to the back-end. As with other design interviews, more senior candidates will face higher expectations for design efficiency and completeness. Even as a junior developer, though, you may be subject to an API design interview.

Understanding the API Design Interview

If you're applying for a data science, machine learning, advanced analytics or artificial intelligence role that has anything to do with designing and building new products, there is a chance you will face an API design interview. Regardless of your API development knowledge, you'll need to assess, plan, design, and review an API capable of requirements set forth by your interviewer.

It's important to remember that these interviews go by quickly too. You'll need to send hirable signals to your interviewers in only around 40 minutes. You'll be expected to drive this conversation on your API design, so being knowledgeable about the general software development lifecycle of an API is crucial.

Let's briefly cover what a successful interview should look like.

1. The interview begins with a very broad problem statement from the interviewer. Usually something along the lines of "Design an API for…"

2. To ensure you understand the problem, you'll need to clarify the functional and non-functional requirements. What will the API need to be able to do? How many users will it have? How is it expected to grow or change?

Try to get all the info you need from the interviewer. (We'll review some necessary details to cover a little later).

3. Next, start the design phase. Here you must address the architecture and protocols needed to satisfy the functional requirements. What endpoints will the API need? What type of data format are you using to exchange info?

4. Finally, you'll need to optimize your design, and consider techniques like client-side rendering and server-side rendering.

Strong hires will not only walk through the entire lifecycle of an API call, but also be able to optimize their design given a latency budget. Given the initial assessments based on the data and usage requirements, consider how you can improve the API — all while staying within a budget of a few hundred milliseconds.

Data Science and the API Design Process

This is a lot to consider, so let's break down the information that you need to convey in the interview. Luckily, we have an acronym for memorization purposes: REDCAMEL.

This acronym stands for:

Requirements (RE) Design Considerations (DC) API Model (AM) Evaluation and Latency Budget (EL).

Requirements (RE) Functional and non-functional requirements should be noted and clarified.

Make the problem statement and motivation for developing the API clear. Following the discussion of requirements, you should make informed back-of-the-envelope calculations on data pipelines, users, and number of requests.

Design Considerations (DC) Next, we can consider the high-level design of the API.

Workflow: How does the API fit into the larger system? API architecture styles: What style API is needed given the client, API gateway, and backend servers? Common API architectures include: REST, GraphQL, and RPC Data formats: How should the data delivered by the API be formatted? Common formats are: JSON, XML, binary Protocols and versions: As a follow-up to the architectural style, what client-server communication protocol will be used?

API Model (AM) This is where you'll first develop a base URL and the necessary API endpoints. Following that, your design should include:

Data entities: What data elements will your API need access to given the use-cases? Request-response type format: What is the format of each operation? Failed requests: Your API should have clearly defined failed request status codes.

Evaluation and Latency Budget (EL) After outlining a functional API structure, it is time to revisit how the design will successfully meet the non-functional requirements.

Evaluate and improve: Navigate the tradeoffs given your design choices. Explain the thought process behind the choices you made. If there are other, more optimal options feel free to suggest changes. Latency budget: Revisit the calculations you made at the beginning. Use these numbers to calculate proposed response times of the API. With this understanding, you can make informed decisions on optimizations and improvements, ensuring your API meets the desired performance goals.

By preparing yourself for the API design interview, understanding the expectations, and familiarizing yourself with the API design process, you will be well-equipped to impress your interviewers and secure the job you desire. As you work in data science, machine learning, artificial intelligence, or advanced analytics, remember that you are more ready than you may think. Everyone has to start someplace, and by following this guide, you're on your way to success.

Conclusion