Data Science Career Roadmap: Math, SQL, Python and Projects
At the 2024 AI Connect K-12 Data Science Conference, SalesLoft senior data scientist Disha An outlined how to prepare for a data science career. Her roadmap covered math and statistics, SQL and Python, domain knowledge, portfolio projects and networking.
By The AI Enterprise Magazine Team4 min read
Key takeaways
- Three pillars. Data science combines math and statistics, computer science and domain knowledge.
- SQL first, then Python. An called SQL the first and most important skill and Python her daily language.
- Fill gaps online. She recommended Coursera, Udemy, edX and Khan Academy, plus key textbooks and research papers.
- Learn the business domain. Examples from finance, oil and gas, and sales showed how context shapes solutions.
- Build and share projects. Kaggle projects, a portfolio and networking can lead to an internship.
At the 2024 AI Connect K-12 Data Science Conference, a virtual event held July 25, 2024, Disha An, Senior Data Scientist at SalesLoft, laid out a step-by-step plan for preparing for a career in data science. She stepped in to speak earlier than scheduled after another speaker had technical trouble.
An said she works at SalesLoft, a sales engagement platform, and previously worked as an applied data scientist at Shell, starting there as a contractor in 2018. She holds a master's degree in financial engineering, where she studied machine learning and deep learning in finance, and said that when she started in 2018 there were few data science degree programs, while today many master's and bachelor's programs exist.
What skills does data science combine?
An framed data science as the meeting of three areas: mathematics and statistics, computer science, and domain knowledge. Each pairing has a name, she said. Computer science plus math and statistics gives artificial intelligence. Math and statistics plus domain knowledge is traditional research, such as the A/B testing and survival analysis pharmaceutical companies run in drug trials. Computer science plus domain knowledge is software development, with Amazon as her example.
Data science, she said, covers all three. She then took them one at a time.
What math and statistics should aspiring data scientists learn?
An listed linear algebra, calculus and gradient descent, probability distributions, statistical significance, Bayes' theorem and hypothesis testing. Most of this is covered in a bachelor's degree in computer science or mathematics, she said.
For gaps, she recommended online learning. She described Coursera as an intro-level equivalent to bachelor's coursework, Udemy as a place where industry experts teach courses with real-world projects, and edX as the more advanced option, with university-level courses. She also pointed to Khan Academy for statistics and calculus basics.
Two reading suggestions stood out: the textbook "An Introduction to Statistical Learning," which she called fundamental, and the research paper "Attention Is All You Need," a foundational paper for generative AI. She advised using sites such as arXiv to keep up with new research.
What computer science and tools matter most?
"The first and most important is SQL," An said. Most companies store data in SQL databases, and data scientists write queries to combine tables and prepare data.
Python came next. She said it is her day-to-day language and that libraries such as TensorFlow, PyTorch and scikit-learn for machine learning and deep learning are written in it. Her list also included:
- Natural language processing basics, such as processing text into tokens
- Extract, transform, load (ETL) work using pandas and NumPy
- Visualization with Matplotlib and Plotly, plus Tableau and Power BI where companies use them
- Spark and Hadoop, including PySpark, for data too large for regular Python
- A general familiarity with cloud platforms from Google, Microsoft and Amazon
- Git for version control when collaborating with data and machine learning engineers
For practice, she recommended HackerRank and LeetCode, noting that LeetCode users share technical interview questions. She said software engineering is the area she personally needs to strengthen, so that her projects can be handed to software engineers for production.
Why does domain knowledge matter?
An defined domain knowledge as background knowledge of the field where data science methods are applied. She gave three examples from her own career. In financial services, understand fraud detection and algorithmic trading. In oil and gas, understand predictive maintenance; at Shell, she said, teams used image processing to detect workers not wearing helmets. In sales, understand revenue prediction.
Knowing the domain, she said, lets a data scientist match an algorithm to the business problem.
How can students build a portfolio and network?
An advised doing projects, and described Kaggle as a platform where users upload free data sets and others share analysis code. She noted that Kaggle also hosts competitions and mentioned an $80,000 prize on a skin cancer detection challenge. She named MachineHack and DrivenData as similar sites.
Once projects are done, she said, build a portfolio and "proactively promote what you've learned" through LinkedIn or Medium posts. She also recommended attending virtual conferences, which can lead to networking and, eventually, an internship.
More ways to learn
AI Connect has 24 free live webinars on its 2027 schedule, in four series: AI × Enterprise Panel Series, AI Demo Days, Agentic AI Showcases and Plug & Play Series. Registration for 2027 events is coming soon; see /webinars/.
Frequently asked questions
What should I learn to become a data scientist?
An named math and statistics, SQL, Python, data wrangling and visualization, big data tools like Spark, cloud basics, Git and domain knowledge.
Where can I practice data science skills?
She recommended Kaggle for projects and competitions, MachineHack and DrivenData, and HackerRank and LeetCode for coding practice.
Is domain knowledge important in data science?
Yes. An said knowing a field, such as fraud detection in finance or predictive maintenance in oil and gas, helps match algorithms to business problems.
This article is based on remarks by Disha An, Senior Data Scientist, SalesLoft at AI Connect K-12 Data Science Conference 2024 on July 25, 2024. More coverage: More AI Connect Events.
Filed under: People, Culture & Leadership, AI Use Cases & Applications
