What does it mean to be a Data Scientist at Jellysmack?
The Data department
At Jellysmack, the Data department comprises 22 people (expected to double this year!) and is composed of three teams : the Data Science team, the Data Analysis team and the Business Analysis team.
The Data Science team which I’m part of, consists of 10 people with various profiles working on different data science problematics, including Deep Learning, Computer Vision and NLP.
Data and data
We have a lot of data at Jellysmack (50B+ rows, 700M+ videos +6% by month) which is really exciting for exploratory data analysis purposes. On the other hand, it means that we need to be careful and analyze the trade-off between accuracy and speed of our solutions. In order to achieve that, we keep learning new techniques to optimize our code and benchmarking state-of-the-art solutions. We also work with data from multiple platforms (e.g., Facebook, Youtube, Instagram, etc.) so we have to stay focused on cross-platform solutions, i.e., our algorithms on Facebook don’t need to be changed when applied to Youtube.
To let our end users become more productive with our data products, we always seek the most intuitive solutions.
Training a complex machine learning model sounds cool for us data scientists, but it’s useless if it fails to meet users’ expectations.
A typical working day
There’s no micro-management, you work at your own pace. So first, to start my working day, I go to Jira to review my tasks and priorities. After identifying the most important tasks, I block appropriate time slots in my daily and weekly calendar. Then I enjoy my day.
There are several types of tasks:
- POC (Proof Of Concept): prototype new solutions or adapt existing solutions to solve a new business problem. In my latest NLP project for example (simply called the “Topic Suggestion”), I analyse video metrics, keywords and tags to recommend promising topics to our social media influencers (e.g. youtubers).
- Industrialization: transform the POC into a data job which will be executed and orchestrated by Airflow. Sometimes we also need to make our POCs available through an API.
- Bugfix: solve problems on my data jobs / APIs in production.
- Organization: write project documentation for technical users (data scientists, developers) and non-technical users (project owners, business people), discuss with the infrastructure/data acquisition teams in order to get the required data for my projects, discuss with the Head of Data to decide which technical solutions are worthy to prototype and discuss with my project team about our shared tasks.
- Personal R&D: I try to spend about 30 minutes per working day to keep myself up to date with the latest libraries, research articles, and new ideas.
- Have fun: from time to time, I go to the Slack room “tech-troll” to have good laughs. For example:
A long-term investment
At Jellysmack we have dedicated time to improve our skills. This is crucial because you can’t solve complex problems with obsolete methods.
Each week we have 5 working hours to improve our skills in development (e.g., code review, pair programming, learn how to use new frameworks/libraries, etc.) and algorithmic (e.g., read research papers, study new machine learning models, brainstorm ideas for current/long-term data science projects). Then we present our work to share what we learned during those hours. A 3.5 hours “tech party” is also organized every 2 weeks to allow everyone to work on their own projects.
To sum up, we are investing 20% of our working hours (13.5 hours over 2 weeks) for the long-term R&D. I’d say this is an impressive number.
Join us
I hope you are now convinced that Jellysmack would be an interesting choice for your career! We are actively looking for (way more) new talents to work with us in Data Science but also in Data and Business Analysis so don’t hesitate, come join us!
All our offers are here ➡ https://jobs.jellysmack.com/