What it takes to be a Machine Learning Engineer at Hurb
The skills and background necessary for being a Machine Learning Engineer at Hurb may surprise you.
Machine learning engineers
Machine learning engineers are responsible for researching, building, and designing models as well as maintaining and improving machine learning systems.
One may think that to be a machine learning engineer you need to have a programming background with many years of experience inside the field, with also strong math, such as algebra, and statistics knowledge.
The truth is that this is a misconception since no human being can have all this knowledge. Furthermore, the pursuit of it would only bring anxiety and frustration. Machine learning engineers need to have a strong foundation on those topics and be self-taught enough to learn throughout the project development process and curious enough that we want to continuously learn since this field is dynamic and fast-changing.
Until recently, there weren’t undergraduate and graduate programs specifically for data science. Most machine learning engineers and data scientists currently in the market don’t have a specific background. We were just curious and wanted to learn about it. That is powerful since it gives a unique opportunity to have diversity.
Also, it is a career full of opportunities, such as:
- Work remotely from anywhere in the world for companies anywhere in the world as well
- Work on totally different projects, such as classification, clustering, time-series, reinforcement learning, image recognition, operational research, and more…
- Working in different sectors such as travel, finance, healthcare, agriculture, and others, there is always a place to improve processes and results using machine learning.
- Strong community to learn and engage with.
It is a fast-growing career. New algorithms are being developed every day; new startups are developing more efficient ways to deploy models to production; low code products and automl are being developed and improved so machine learning engineers can focus more on how to solve the business problem instead of spending most of the time coding everything from scratch. Consequently, companies are becoming more dependent on machine learning models and leveraging their power to improve service and processes, as well as investing in companies that are trying to solve business problems using ML.
Machine learning team at Hurb
We have machine learning engineers from various backgrounds, such as production engineering, electrical engineering, mechanical engineering, environmental engineering, biotechnology technician, computer science, and more. All have different perspectives and add value in their own way.
Our skills can be complemented, for instance, production engineers are strong at recognizing business needs and how to address the problem in the best way to generate value as soon as possible, focusing on statistics and modeling, but the programming part can be challenging. That is when people with a computer science or software engineering background can help. Therefore, we build a team with complementary skills.
This is especially beneficial for us since we deal with different projects that require different skills, such as:
- Predict if the customer will accept or refuse a flight option in order to reduce rework and increase customer satisfaction, you can learn more about this project at our medium article How does Hurb use Machine Learning to choose the best flight options for the customers.
- Clustering hotels for e-commerce improvements.
- Airfare forecast for the next 12 months to help us better pricing our packages, negotiate with airline companies, predict our cash flow and more…
- Select the best hotel images to display on our website to improve customer experience.
Moreover, we want the machine learning engineers to have full autonomy of the whole machine learning life cycle, from the business evaluation to the productization and monitoring. This way, we can reduce the project’s development time, rework, and communication failure risks since we are not dependent on other teams and their prioritization and knowledge of the model.
In order to achieve this autonomy, we could follow two paths: build a team of software and machine learning unicorn experts or create abstractions that reduce the task’s complexity. At Hurb, we opted for the second approach. We have two roles of machine learning engineers, solutions and platform. The Solutions are the ones we have talked about so far that work together with other Hurb’s areas to develop ML solutions. In contrast, Platform are the ones focusing on improving our machine learning stack, creating an MLOps platform that abstracts most of the technical complexity of software development in order to develop quality and efficient data products, following best practices and so solutions machine learning engineers don’t need to be software experts.
It is important to highlight that this approach allows us to have diversity and focus on technical skills and highly valuing soft skills. We all need to deal with stakeholders, Solution’s machine learning engineers need to communicate efficiently with product managers, developers, directors, and others. In order to do that, we need to understand the business, communicate the model’s results, and document the project in a simple way so everyone can understand. At the Platform’s machine learning engineer side, communication is also essential to understand the data team's pain points and how to solve them properly. In addition, it allows us to have a great environment within the team, working together to deliver high-value data products to Hurb.
In conclusion, we leverage the power of diversity and complementary roles as well as technologies, such as BentoML for serving, Flyte for orchestration, and MLFlow for tracking and model versioning. In addition, we are taking advantage of Google’s stack and developing an internal library called Hurbert to improve our processes.
Nonetheless, our data team Is supported by data engineers responsible for data governance, quality, and ingestion. Also, we can count on data analysts that are responsible for generating value through data analysis and have deep business knowledge. Both teams help us through all model development and are essential to the whole process. If you are interested in more detail regarding our data engineering platform, feel free to check our medium post Data Platform Architecture at Hurb.
Furthermore, Hurb is a travel tech company that is highly investing in machine learning and its team. We doubled the team in 2022, with all members working hybrid or fully remotely, which allowed us to increase diversity by hiring not only from Rio de Janeiro, but from different regions of Brazil, such as Bahia, Rio Grande do Norte, Rio Grande do Sul, as well as abroad, from France. In addition, we have the opportunity to attend different events around the world. For instance, in 2022, we participated at MLOps World and Collision in Toronto, Google’s Advanced Solutions Lab in New York, The Developers Conference (TDC) in São Paulo, and more. We are also highly connected with the machine learning community, engaging in hackathons, contributing to open-source projects, and more. If you are interested, you can get to know more about some of our machine learning team members on our youtube channel Life at Hurb.
Hope you enjoyed this article! If you want to join our machine learning team, find our job openings here. Also, you can check out more about our work at Hurb on our Medium page.
